Authors: Amazon AGI (JC), Aaron Langford (JC), Aayush Shah (JC), Abhanshu Gupta (JC), Abhimanyu Bhatter (JC), Abhinav Goyal (JC), Abhinav Mathur (JC), Abhinav Mohanty (JC), Abhishek Kumar (JC), Abhishek Sethi (JC), Abi Komma (JC), Abner Pena (JC), Achin Jain (JC), Adam Kunysz (JC), Adam Opyrchal (JC), Adarsh Singh (JC), Aditya Rawal (JC), Adok Achar Budihal Prasad (JC), Adri\`a de Gispert (JC), Agnika Kumar (JC), Aishwarya Aryamane (JC), Ajay Nair (JC), Akilan M (JC), Akshaya Iyengar (JC), Akshaya Vishnu Kudlu Shanbhogue (JC), Alan He (JC), Alessandra Cervone (JC), Alex Loeb (JC), Alex Zhang (JC), Alexander Fu (JC), Alexander Lisnichenko (JC), Alexander Zhipa (JC), Alexandros Potamianos (JC), Ali Kebarighotbi (JC), Aliakbar Daronkolaei (JC), Alok Parmesh (JC), Amanjot Kaur Samra (JC), Ameen Khan (JC), Amer Rez (JC), Amir Saffari (JC), Amit Agarwalla (JC), Amit Jhindal (JC), Amith Mamidala (JC), Ammar Asmro (JC), Amulya Ballakur (JC), Anand Mishra (JC), Anand Sridharan (JC), Anastasiia Dubinina (JC), Andre Lenz (JC), Andreas Doerr (JC), Andrew Keating (JC), Andrew Leaver (JC), Andrew Smith (JC), Andrew Wirth (JC), Andy Davey (JC), Andy Rosenbaum (JC), Andy Sohn (JC), Angela Chan (JC), Aniket Chakrabarti (JC), Anil Ramakrishna (JC), Anirban Roy (JC), Anita Iyer (JC), Anjali Narayan-Chen (JC), Ankith Yennu (JC), Anna Dabrowska (JC), Anna Gawlowska (JC), Anna Rumshisky (JC), Anna Turek (JC), Anoop Deoras (JC), Anton Bezruchkin (JC), Anup Prasad (JC), Anupam Dewan (JC), Anwith Kiran (JC), Apoorv Gupta (JC), Aram Galstyan (JC), Aravind Manoharan (JC), Arijit Biswas (JC), Arindam Mandal (JC), Arpit Gupta (JC), Arsamkhan Pathan (JC), Arun Nagarajan (JC), Arushan Rajasekaram (JC), Arvind Sundararajan (JC), Ashwin Ganesan (JC), Ashwin Swaminathan (JC), Athanasios Mouchtaris (JC), Audrey Champeau (JC), Avik Ray (JC), Ayush Jaiswal (JC), Ayush Sharma (JC), Bailey Keefer (JC), Balamurugan Muthiah (JC), Beatriz Leon-Millan (JC), Ben Koopman (JC), Ben Li (JC), Benjamin Biggs (JC), Benjamin Ott (JC), Bhanu Vinzamuri (JC), Bharath Venkatesh (JC), Bhavana Ganesh (JC), Bhoomit Vasani (JC), Bill Byrne (JC), Bill Hsu (JC), Bincheng Wang (JC), Blake King (JC), Blazej Gorny (JC), Bo Feng (JC), Bo Zheng (JC), Bodhisattwa Paul (JC), Bofan Sun (JC), Bofeng Luo (JC), Bowen Chen (JC), Bowen Xie (JC), Boya Yu (JC), Brendan Jugan (JC), Brett Panosh (JC), Brian Collins (JC), Brian Thompson (JC), Can Karakus (JC), Can Liu (JC), Carl Lambrecht (JC), Carly Lin (JC), Carolyn Wang (JC), Carrie Yuan (JC), Casey Loyda (JC), Cezary Walczak (JC), Chalapathi Choppa (JC), Chandana Satya Prakash (JC), Chankrisna Richy Meas (JC), Charith Peris (JC), Charles Recaido (JC), Charlie Xu (JC), Charul Sharma (JC), Chase Kernan (JC), Chayut Thanapirom (JC), Chengwei Su (JC), Chenhao Xu (JC), Chenhao Yin (JC), Chentao Ye (JC), Chenyang Tao (JC), Chethan Parameshwara (JC), Ching-Yun Chang (JC), Chong Li (JC), Chris Hench (JC), Chris Tran (JC), Christophe Dupuy (JC), Christopher Davis (JC), Christopher DiPersio (JC), Christos Christodoulopoulos (JC), Christy Li (JC), Chun Chen (JC), Claudio Delli Bovi (JC), Clement Chung (JC), Cole Hawkins (JC), Connor Harris (JC), Corey Ropell (JC), Cynthia He (JC), DK Joo (JC), Dae Yon Hwang (JC), Dan Rosen (JC), Daniel Elkind (JC), Daniel Pressel (JC), Daniel Zhang (JC), Danielle Kimball (JC), Daniil Sorokin (JC), Dave Goodell (JC), Davide Modolo (JC), Dawei Zhu (JC), Deepikaa Suresh (JC), Deepti Ragha (JC), Denis Filimonov (JC), Denis Foo Kune (JC), Denis Romasanta Rodriguez (JC), Devamanyu Hazarika (JC), Dhananjay Ram (JC), Dhawal Parkar (JC), Dhawal Patel (JC), Dhwanil Desai (JC), Dinesh Singh Rajput (JC), Disha Sule (JC), Diwakar Singh (JC), Dmitriy Genzel (JC), Dolly Goldenberg (JC), Dongyi He (JC), Dumitru Hanciu (JC), Dushan Tharmal (JC), Dzmitry Siankovich (JC), Edi Cikovic (JC), Edwin Abraham (JC), Ekraam Sabir (JC), Elliott Olson (JC), Emmett Steven (JC), Emre Barut (JC), Eric Jackson (JC), Ethan Wu (JC), Evelyn Chen (JC), Ezhilan Mahalingam (JC), Fabian Triefenbach (JC), Fan Yang (JC), Fangyu Liu (JC), Fanzi Wu (JC), Faraz Tavakoli (JC), Farhad Khozeimeh (JC), Feiyang Niu (JC), Felix Hieber (JC), Feng Li (JC), Firat Elbey (JC), Florian Krebs (JC), Florian Saupe (JC), Florian Spr\"unken (JC), Frank Fan (JC), Furqan Khan (JC), Gabriela De Vincenzo (JC), Gagandeep Kang (JC), George Ding (JC), George He (JC), George Yeung (JC), Ghada Qaddoumi (JC), Giannis Karamanolakis (JC), Goeric Huybrechts (JC), Gokul Maddali (JC), Gonzalo Iglesias (JC), Gordon McShane (JC), Gozde Sahin (JC), Guangtai Huang (JC), Gukyeong Kwon (JC), Gunnar A. Sigurdsson (JC), Gurpreet Chadha (JC), Gururaj Kosuru (JC), Hagen Fuerstenau (JC), Hah Hah (JC), Haja Maideen (JC), Hajime Hosokawa (JC), Han Liu (JC), Han-Kai Hsu (JC), Hann Wang (JC), Hao Li (JC), Hao Yang (JC), Haofeng Zhu (JC), Haozheng Fan (JC), Harman Singh (JC), Harshavardhan Kaluvala (JC), Hashim Saeed (JC), He Xie (JC), Helian Feng (JC), Hendrix Luo (JC), Hengzhi Pei (JC), Henrik Nielsen (JC), Hesam Ilati (JC), Himanshu Patel (JC), Hongshan Li (JC), Hongzhou Lin (JC), Hussain Raza (JC), Ian Cullinan (JC), Imre Kiss (JC), Inbarasan Thangamani (JC), Indrayani Fadnavis (JC), Ionut Teodor Sorodoc (JC), Irem Ertuerk (JC), Iryna Yemialyanava (JC), Ishan Soni (JC), Ismail Jelal (JC), Ivan Tse (JC), Jack FitzGerald (JC), Jack Zhao (JC), Jackson Rothgeb (JC), Jacky Lee (JC), Jake Jung (JC), Jakub Debski (JC), Jakub Tomczak (JC), James Jeun (JC), James Sanders (JC), Jason Crowley (JC), Jay Lee (JC), Jayakrishna Anvesh Paidy (JC), Jayant Tiwari (JC), Jean Farmer (JC), Jeff Solinsky (JC), Jenna Lau (JC), Jeremy Savareese (JC), Jerzy Zagorski (JC), Ji Dai (JC), Jiacheng (JC), Gu (Skyler), Jiahui Li (Skyler), Jian (Skyler), Zheng (QZ), Jianhua Lu (QZ), Jianhua Wang (QZ), Jiawei Dai (QZ), Jiawei Mo (QZ), Jiaxi Xu (QZ), Jie Liang (QZ), Jie Yang (QZ), Jim Logan (QZ), Jimit Majmudar (QZ), Jing Liu (QZ), Jinghong Miao (QZ), Jingru Yi (QZ), Jingyang Jin (QZ), Jiun-Yu Kao (QZ), Jixuan Wang (QZ), Jiyang Wang (QZ), Joe Pemberton (QZ), Joel Carlson (QZ), Joey Blundell (QZ), John Chin-Jew (QZ), John He (QZ), Jonathan Ho (QZ), Jonathan Hueser (QZ), Jonathan Lunt (QZ), Jooyoung Lee (QZ), Joshua Tan (QZ), Joyjit Chatterjee (QZ), Judith Gaspers (QZ), Jue Wang (QZ), Jun Fang (QZ), Jun Tang (QZ), Jun Wan (QZ), Jun Wu (QZ), Junlei Wang (QZ), Junyi Shi (QZ), Justin Chiu (QZ), Justin Satriano (QZ), Justin Yee (QZ), Jwala Dhamala (QZ), Jyoti Bansal (QZ), Kai Zhen (QZ), Kai-Wei Chang (QZ), Kaixiang Lin (QZ), Kalyan Raman (QZ), Kanthashree Mysore Sathyendra (QZ), Karabo Moroe (QZ), Karan Bhandarkar (QZ), Karan Kothari (QZ), Karolina Owczarzak (QZ), Karthick Gopalswamy (QZ), Karthick Ravi (QZ), Karthik Ramakrishnan (QZ), Karthika Arumugam (QZ), Kartik Mehta (QZ), Katarzyna Konczalska (QZ), Kavya Ravikumar (QZ), Ke Tran (QZ), Kechen Qin (QZ), Kelin Li (QZ), Kelvin Li (QZ), Ketan Kulkarni (QZ), Kevin Angelo Rodrigues (QZ), Keyur Patel (QZ), Khadige Abboud (QZ), Kiana Hajebi (QZ), Klaus Reiter (QZ), Kris Schultz (QZ), Krishna Anisetty (QZ), Krishna Kotnana (QZ), Kristen Li (QZ), Kruthi Channamallikarjuna (QZ), Krzysztof Jakubczyk (QZ), Kuba Pierewoj (QZ), Kunal Pal (QZ), Kunwar Srivastav (QZ), Kyle Bannerman (QZ), Lahari Poddar (QZ), Lakshmi Prasad (QZ), Larry Tseng (QZ), Laxmikant Naik (QZ), Leena Chennuru Vankadara (QZ), Lenon Minorics (QZ), Leo Liu (QZ), Leonard Lausen (QZ), Leonardo F. R. Ribeiro (QZ), Li Zhang (QZ), Lili Gehorsam (QZ), Ling Qi (QZ), Lisa Bauer (QZ), Lori Knapp (QZ), Lu Zeng (QZ), Lucas Tong (QZ), Lulu Wong (QZ), Luoxin Chen (QZ), Maciej Rudnicki (QZ), Mahdi Namazifar (QZ), Mahesh Jaliminche (QZ), Maira Ladeira Tanke (QZ), Manasi Gupta (QZ), Mandeep Ahlawat (QZ), Mani Khanuja (QZ), Mani Sundaram (QZ), Marcin Leyk (QZ), Mariusz Momotko (QZ), Markus Boese (QZ), Markus Dreyer (QZ), Markus Mueller (QZ), Mason Fu (QZ), Mateusz G\'orski (QZ), Mateusz Mastalerczyk (QZ), Matias Mora (QZ), Matt Johnson (QZ), Matt Scott (QZ), Matthew Wen (QZ), Max Barysau (QZ), Maya Boumerdassi (QZ), Maya Krishnan (QZ), Mayank Gupta (QZ), Mayank Hirani (QZ), Mayank Kulkarni (QZ), Meganathan Narayanasamy (QZ), Melanie Bradford (QZ), Melanie Gens (QZ), Melissa Burke (QZ), Meng Jin (QZ), Miao Chen (QZ), Michael Denkowski (QZ), Michael Heymel (QZ), Michael Krestyaninov (QZ), Michal Obirek (QZ), Michalina Wichorowska (QZ), Micha{\l} Miotk (QZ), Milosz Watroba (QZ), Mingyi Hong (QZ), Mingzhi Yu (QZ), Miranda Liu (QZ), Mohamed Gouda (QZ), Mohammad El-Shabani (QZ), Mohammad Ghavamzadeh (QZ), Mohit Bansal (QZ), Morteza Ziyadi (QZ), Nan Xia (QZ), Nathan Susanj (QZ), Nav Bhasin (QZ), Neha Goswami (QZ), Nehal Belgamwar (QZ), Nicolas Anastassacos (QZ), Nicolas Bergeron (QZ), Nidhi Jain (QZ), Nihal Jain (QZ), Niharika Chopparapu (QZ), Nik Xu (QZ), Nikko Strom (QZ), Nikolaos Malandrakis (QZ), Nimisha Mishra (QZ), Ninad Parkhi (QZ), Ninareh Mehrabi (QZ), Nishita Sant (QZ), Nishtha Gupta (QZ), Nitesh Sekhar (QZ), Nithin Rajeev (QZ), Nithish Raja Chidambaram (QZ), Nitish Dhar (QZ), Noor Bhagwagar (QZ), Noy Konforty (QZ), Omar Babu (QZ), Omid Razavi (QZ), Orchid Majumder (QZ), Osama Dar (QZ), Oscar Hsu (QZ), Pablo Kvitca (QZ), Pallavi Pandey (QZ), Parker Seegmiller (QZ), Patrick Lange (QZ), Paul Ferraro (QZ), Payal Motwani (QZ), Pegah Kharazmi (QZ), Pei Wang (QZ), Pengfei Liu (QZ), Peter Bradtke (QZ), Peter G\"otz (QZ), Peter Zhou (QZ), Pichao Wang (QZ), Piotr Poskart (QZ), Pooja Sonawane (QZ), Pradeep Natarajan (QZ), Pradyun Ramadorai (QZ), Pralam Shah (QZ), Prasad Nirantar (QZ), Prasanthi Chavali (QZ), Prashan Wanigasekara (QZ), Prashant Saraf (QZ), Prashun Dey (QZ), Pratyush Pant (QZ), Prerak Pradhan (QZ), Preyaa Patel (QZ), Priyanka Dadlani (QZ), Prudhvee Narasimha Sadha (QZ), Qi Dong (QZ), Qian Hu (QZ), Qiaozi (QZ), Gao (Sean), Qing Liu (Sean), Quinn Lam (Sean), Quynh Do (Sean), R. Manmatha (Sean), Rachel Willis (Sean), Rafael Liu (Sean), Rafal Ellert (Sean), Rafal Kalinski (Sean), Rafi Al Attrach (Sean), Ragha Prasad (Sean), Ragini Prasad (Sean), Raguvir Kunani (Sean), Rahul Gupta (Sean), Rahul Sharma (Sean), Rahul Tewari (Sean), Rajaganesh Baskaran (Sean), Rajan Singh (Sean), Rajiv Gupta (Sean), Rajiv Reddy (Sean), Rajshekhar Das (Sean), Rakesh Chada (Sean), Rakesh Vaideeswaran Mahesh (Sean), Ram Chandrasekaran (Sean), Ramesh Nallapati (Sean), Ran Xue (Sean), Rashmi Gangadharaiah (Sean), Ravi Rachakonda (Sean), Renxian Zhang (Sean), Rexhina Blloshmi (Sean), Rishabh Agrawal (Sean), Robert Enyedi (Sean), Robert Lowe (Sean), Robik Shrestha (Sean), Robinson Piramuthu (Sean), Rohail Asad (Sean), Rohan Khanna (Sean), Rohan Mukherjee (Sean), Rohit Mittal (Sean), Rohit Prasad (Sean), Rohith Mysore Vijaya Kumar (Sean), Ron Diamant (Sean), Ruchita Gupta (Sean), Ruiwen Li (Sean), Ruoying Li (Sean), Rushabh Fegade (Sean), Ruxu Zhang (Sean), Ryan Arbow (Sean), Ryan Chen (Sean), Ryan Gabbard (Sean), Ryan Hoium (Sean), Ryan King (Sean), Sabarishkumar Iyer (Sean), Sachal Malick (Sean), Sahar Movaghati (Sean), Sai Balakavi (Sean), Sai Jakka (Sean), Sai Kashyap Paruvelli (Sean), Sai Muralidhar Jayanthi (Sean), Saicharan Shriram Mujumdar (Sean), Sainyam Kapoor (Sean), Sajjad Beygi (Sean), Saket Dingliwal (Sean), Saleh Soltan (Sean), Sam Ricklin (Sean), Sam Tucker (Sean), Sameer Sinha (Sean), Samridhi Choudhary (Sean), Samson Tan (Sean), Samuel Broscheit (Sean), Samuel Schulter (Sean), Sanchit Agarwal (Sean), Sandeep Atluri (Sean), Sander Valstar (Sean), Sanjana Shankar (Sean), Sanyukta Sanyukta (Sean), Sarthak Khanna (Sean), Sarvpriye Khetrapal (Sean), Satish Janakiraman (Sean), Saumil Shah (Sean), Saurabh Akolkar (Sean), Saurabh Giri (Sean), Saurabh Khandelwal (Sean), Saurabh Pawar (Sean), Saurabh Sahu (Sean), Sean Huang (Sean), Sejun Ra (Sean), Senthilkumar Gopal (Sean), Sergei Dobroshinsky (Sean), Shadi Saba (Sean), Shamik Roy (Sean), Shamit Lal (Sean), Shankar Ananthakrishnan (Sean), Sharon Li (Sean), Shashwat Srijan (Sean), Shekhar Bhide (Sean), Sheng Long Tang (Sean), Sheng Zha (Sean), Shereen Oraby (Sean), Sherif Mostafa (Sean), Shiqi Li (Sean), Shishir Bharathi (Sean), Shivam Prakash (Sean), Shiyuan Huang (Sean), Shreya Yembarwar (Sean), Shreyas Pansare (Sean), Shreyas Subramanian (Sean), Shrijeet Joshi (Sean), Shuai Liu (Sean), Shuai Tang (Sean), Shubham Chandak (Sean), Shubham Garg (Sean), Shubham Katiyar (Sean), Shubham Mehta (Sean), Shubham Srivastav (Sean), Shuo Yang (Sean), Siddalingesha D S (Sean), Siddharth Choudhary (Sean), Siddharth Singh Senger (Sean), Simon Babb (Sean), Sina Moeini (Sean), Siqi Deng (Sean), Siva Loganathan (Sean), Slawomir Domagala (Sean), Sneha Narkar (Sean), Sneha Wadhwa (Sean), Songyang Zhang (Sean), Songyao Jiang (Sean), Sony Trenous (Sean), Soumajyoti Sarkar (Sean), Soumya Saha (Sean), Sourabh Reddy (Sean), Sourav Dokania (Sean), Spurthideepika Sandiri (Sean), Spyros Matsoukas (Sean), Sravan Bodapati (Sean), Sri Harsha Reddy Wdaru (Sean), Sridevi Yagati Venkateshdatta (Sean), Srikanth Ronanki (Sean), Srinivasan R Veeravanallur (Sean), Sriram Venkatapathy (Sean), Sriramprabhu Sankaraguru (Sean), Sruthi Gorantla (Sean), Sruthi Karuturi (Sean), Stefan Schroedl (Sean), Subendhu Rongali (Sean), Subhasis Kundu (Sean), Suhaila Shakiah (Sean), Sukriti Tiwari (Sean), Sumit Bharti (Sean), Sumita Sami (Sean), Sumith Mathew (Sean), Sunny Yu (Sean), Sunwoo Kim (Sean), Suraj Bajirao Malode (Sean), Susana Cumplido Riel (Sean), Swapnil Palod (Sean), Swastik Roy (Sean), Syed Furqhan (Sean), Tagyoung Chung (Sean), Takuma Yoshitani (Sean), Taojiannan Yang (Sean), Tejaswi Chillakura (Sean), Tejwant Bajwa (Sean), Temi Lajumoke (Sean), Thanh Tran (Sean), Thomas Gueudre (Sean), Thomas Jung (Sean), Tianhui Li (Sean), Tim Seemman (Sean), Timothy Leffel (Sean), Tingting Xiang (Sean), Tirth Patel (Sean), Tobias Domhan (Sean), Tobias Falke (Sean), Toby Guo (Sean), Tom Li (Sean), Tomasz Horszczaruk (Sean), Tomasz Jedynak (Sean), Tushar Kulkarni (Sean), Tyst Marin (Sean), Tytus Metrycki (Sean), Tzu-Yen Wang (Sean), Umang Jain (Sean), Upendra Singh (Sean), Utkarsh Chirimar (Sean), Vaibhav Gupta (Sean), Vanshil Shah (Sean), Varad Deshpande (Sean), Varad Gunjal (Sean), Varsha Srikeshava (Sean), Varsha Vivek (Sean), Varun Bharadwaj (Sean), Varun Gangal (Sean), Varun Kumar (Sean), Venkatesh Elango (Sean), Vicente Ordonez (Sean), Victor Soto (Sean), Vignesh Radhakrishnan (Sean), Vihang Patel (Sean), Vikram Singh (Sean), Vinay Varma Kolanuvada (Sean), Vinayshekhar Bannihatti Kumar (Sean), Vincent Auvray (Sean), Vincent Cartillier (Sean), Vincent Ponzo (Sean), Violet Peng (Sean), Vishal Khandelwal (Sean), Vishal Naik (Sean), Vishvesh Sahasrabudhe (Sean), Vitaliy Korolev (Sean), Vivek Gokuladas (Sean), Vivek Madan (Sean), Vivek Subramanian (Sean), Volkan Cevher (Sean), Vrinda Gupta (Sean), Wael Hamza (Sean), Wei Zhang (Sean), Weitong Ruan (Sean), Weiwei Cheng (Sean), Wen Zhang (Sean), Wenbo Zhao (Sean), Wenyan Yao (Sean), Wenzhuo Ouyang (Sean), Wesley Dashner (Sean), William Campbell (Sean), William Lin (Sean), Willian Martin (Sean), Wyatt Pearson (Sean), Xiang Jiang (Sean), Xiangxing Lu (Sean), Xiangyang Shi (Sean), Xianwen Peng (Sean), Xiaofeng Gao (Sean), Xiaoge Jiang (Sean), Xiaohan Fei (Sean), Xiaohui Wang (Sean), Xiaozhou Joey Zhou (Sean), Xin Feng (Sean), Xinyan Zhao (Sean), Xinyao Wang (Sean), Xinyu Li (Sean), Xu Zhang (Sean), Xuan Wang (Sean), Xuandi Fu (Sean), Xueling Yuan (Sean), Xuning Wang (Sean), Yadunandana Rao (Sean), Yair Tavizon (Sean), Yan Rossiytsev (Sean), Yanbei Chen (Sean), Yang Liu (Sean), Yang Zou (Sean), Yangsook Park (Sean), Yannick Versley (Sean), Yanyan Zhang (Sean), Yash Patel (Sean), Yen-Cheng Lu (Sean), Yi Pan (Sean), Yi-Hsiang (Sean), Lai (Rex), Yichen Hu (Rex), Yida Wang (Rex), Yiheng Zhou (Rex), Yilin Xiang (Rex), Ying Shi (Rex), Ying Wang (Rex), Yishai Galatzer (Rex), Yongxin Wang (Rex), Yorick Shen (Rex), Yuchen Sun (Rex), Yudi Purwatama (Rex), Yue (Rex), Wu (Chris), Yue Gu (Chris), Yuechun Wang (Chris), Yujun Zeng (Chris), Yuncong Chen (Chris), Yunke Zhou (Chris), Yusheng Xie (Chris), Yvon Guy (Chris), Zbigniew Ambrozinski (Chris), Zhaowei Cai (Chris), Zhen Zhang (Chris), Zheng Wang (Chris), Zhenghui Jin (Chris), Zhewei Zhao (Chris), Zhiheng Li (Chris), Zhiheng Luo (Chris), Zhikang Zhang (Chris), Zhilin Fang (Chris), Zhiqi Bu (Chris), Zhiyuan Wang (Chris), Zhizhong Li (Chris), Zijian Wang (Chris), Zimeng (Chris), Qiu, Zishi Li
Abstract: We present Amazon Nova, a new generation of state-of-the-art foundation models that deliver frontier intelligence and industry-leading price performance. Amazon Nova Pro is a highly-capable multimodal model with the best combination of accuracy, speed, and cost for a wide range of tasks. Amazon Nova Lite is a low-cost multimodal model that is lightning fast for processing images, video, documents and text. Amazon Nova Micro is a text-only model that delivers our lowest-latency responses at very low cost. Amazon Nova Canvas is an image generation model that creates professional grade images with rich customization controls. Amazon Nova Reel is a video generation model offering high-quality outputs, customization, and motion control. Our models were built responsibly and with a commitment to customer trust, security, and reliability. We report benchmarking results for core capabilities, agentic performance, long context, functional adaptation, runtime performance, and human evaluation.
Authors: Been Kim, John Hewitt, Neel Nanda, Noah Fiedel, Oyvind Tafjord
Abstract: The era of Large Language Models (LLMs) presents a new opportunity for interpretability--agentic interpretability: a multi-turn conversation with an LLM wherein the LLM proactively assists human understanding by developing and leveraging a mental model of the user, which in turn enables humans to develop better mental models of the LLM. Such conversation is a new capability that traditional `inspective' interpretability methods (opening the black-box) do not use. Having a language model that aims to teach and explain--beyond just knowing how to talk--is similar to a teacher whose goal is to teach well, understanding that their success will be measured by the student's comprehension. While agentic interpretability may trade off completeness for interactivity, making it less suitable for high-stakes safety situations with potentially deceptive models, it leverages a cooperative model to discover potentially superhuman concepts that can improve humans' mental model of machines. Agentic interpretability introduces challenges, particularly in evaluation, due to what we call `human-entangled-in-the-loop' nature (humans responses are integral part of the algorithm), making the design and evaluation difficult. We discuss possible solutions and proxy goals. As LLMs approach human parity in many tasks, agentic interpretability's promise is to help humans learn the potentially superhuman concepts of the LLMs, rather than see us fall increasingly far from understanding them.
Authors: Elhoucine Elfatimi, Yassir Lekbach, Swayam Prakash, Lbachir BenMohamed
Abstract: In the past, the development of vaccines and immunotherapeutics relied heavily on trial-and-error experimentation and extensive in vivo testing, often requiring years of pre-clinical and clinical trials. Today, artificial intelligence (AI) and deep learning (DL) are actively transforming vaccine and immunotherapeutic design, by (i) offering predictive frameworks that support rapid, data-driven decision-making; (ii) increasingly being implemented as time- and resource-efficient strategies that integrate computational models, systems vaccinology, and multi-omics data to better phenotype, differentiate, and classify patient diseases and cancers; predict patients' immune responses; and identify the factors contributing to optimal vaccine and immunotherapeutic protective efficacy; (iii) refining the selection of B- and T-cell antigen/epitope targets to enhance efficacy and durability of immune protection; and (iv) enabling a deeper understanding of immune regulation, immune evasion, immune checkpoints, and regulatory pathways. The future of AI and DL points toward (i) replacing animal preclinical testing of drugs, vaccines, and immunotherapeutics with computational-based models, as recently proposed by the United States FDA; and (ii) enabling real-time in vivo modeling for immunobridging and prediction of protection in clinical trials. This may result in a fast and transformative shift for the development of personal vaccines and immunotherapeutics against infectious pathogens and cancers.
Authors: Yujie Zhao, Zhijing Wu, Hejia Zhang, Zhongming Yu, Wentao Ni, Chia-Tung Ho, Haoxing Ren, Jishen Zhao
Abstract: LLM-assisted hardware verification is gaining substantial attention due to its potential to significantly reduce the cost and effort of crafting effective testbenches. It also serves as a critical enabler for LLM-aided end-to-end hardware language design. However, existing current LLMs often struggle with Register Transfer Level (RTL) code generation, resulting in testbenches that exhibit functional errors in Hardware Description Languages (HDL) logic. Motivated by the strong performance of LLMs in Python code generation under inference-time sampling strategies, and their promising capabilities as judge agents, we propose PRO-V a fully program generation multi-agent system for robust RTL verification. Pro-V incorporates an efficient best-of-n iterative sampling strategy to enhance the correctness of generated testbenches. Moreover, it introduces an LLM-as-a-judge aid validation framework featuring an automated prompt generation pipeline. By converting rule-based static analysis from the compiler into natural language through in-context learning, this pipeline enables LLMs to assist the compiler in determining whether verification failures stem from errors in the RTL design or the testbench. PRO-V attains a verification accuracy of 87.17% on golden RTL implementations and 76.28% on RTL mutants. Our code is open-sourced at https://github.com/stable-lab/Pro-V.
Authors: Ren Yi, Octavian Suciu, Adria Gascon, Sarah Meiklejohn, Eugene Bagdasarian, Marco Gruteser
Abstract: We study the ability of language models to reason about appropriate information disclosure - a central aspect of the evolving field of agentic privacy. Whereas previous works have focused on evaluating a model's ability to align with human decisions, we examine the role of ambiguity and missing context on model performance when making information-sharing decisions. We identify context ambiguity as a crucial barrier for high performance in privacy assessments. By designing Camber, a framework for context disambiguation, we show that model-generated decision rationales can reveal ambiguities and that systematically disambiguating context based on these rationales leads to significant accuracy improvements (up to 13.3\% in precision and up to 22.3\% in recall) as well as reductions in prompt sensitivity. Overall, our results indicate that approaches for context disambiguation are a promising way forward to enhance agentic privacy reasoning.
Authors: Cosimo Spera, Garima Agrawal
Abstract: The relationship between humans and artificial intelligence is no longer science fiction -- it's a growing reality reshaping how we live and work. AI has moved beyond research labs into everyday life, powering customer service chats, personalizing travel, aiding doctors in diagnosis, and supporting educators. What makes this moment particularly compelling is AI's increasing collaborative nature. Rather than replacing humans, AI augments our capabilities -- automating routine tasks, enhancing decisions with data, and enabling creativity in fields like design, music, and writing. The future of work is shifting toward AI agents handling tasks autonomously, with humans as supervisors, strategists, and ethical stewards. This flips the traditional model: instead of humans using AI as a tool, intelligent agents will operate independently within constraints, managing everything from scheduling and customer service to complex workflows. Humans will guide and fine-tune these agents to ensure alignment with goals, values, and context. This shift offers major benefits -- greater efficiency, faster decisions, cost savings, and scalability. But it also brings risks: diminished human oversight, algorithmic bias, security flaws, and a widening skills gap. To navigate this transition, organizations must rethink roles, invest in upskilling, embed ethical principles, and promote transparency. This paper examines the technological and organizational changes needed to enable responsible adoption of AI-first systems -- where autonomy is balanced with human intent, oversight, and values.
Authors: Ali Asadi, Krishnendu Chatterjee, Jakob de Raaij
Abstract: Deterministic Markov Decision Processes (DMDPs) are a mathematical framework for decision-making where the outcomes and future possible actions are deterministically determined by the current action taken. DMDPs can be viewed as a finite directed weighted graph, where in each step, the controller chooses an outgoing edge. An objective is a measurable function on runs (or infinite trajectories) of the DMDP, and the value for an objective is the maximal cumulative reward (or weight) that the controller can guarantee. We consider the classical mean-payoff (aka limit-average) objective, which is a basic and fundamental objective. Howard's policy iteration algorithm is a popular method for solving DMDPs with mean-payoff objectives. Although Howard's algorithm performs well in practice, as experimental studies suggested, the best known upper bound is exponential and the current known lower bound is as follows: For the input size $I$, the algorithm requires $\tilde{\Omega}(\sqrt{I})$ iterations, where $\tilde{\Omega}$ hides the poly-logarithmic factors, i.e., the current lower bound on iterations is sub-linear with respect to the input size. Our main result is an improved lower bound for this fundamental algorithm where we show that for the input size $I$, the algorithm requires $\tilde{\Omega}(I)$ iterations.
Authors: Zhenning Yang, Archit Bhatnagar, Yiming Qiu, Tongyuan Miao, Patrick Tser Jern Kon, Yunming Xiao, Yibo Huang, Martin Casado, Ang Chen
Abstract: Cloud infrastructure is the cornerstone of the modern IT industry. However, managing this infrastructure effectively requires considerable manual effort from the DevOps engineering team. We make a case for developing AI agents powered by large language models (LLMs) to automate cloud infrastructure management tasks. In a preliminary study, we investigate the potential for AI agents to use different cloud/user interfaces such as software development kits (SDK), command line interfaces (CLI), Infrastructure-as-Code (IaC) platforms, and web portals. We report takeaways on their effectiveness on different management tasks, and identify research challenges and potential solutions.
Authors: Kehua Chen, Shucheng Zhang, Yinhai Wang
Abstract: Modeling vehicle interactions at unsignalized intersections is a challenging task due to the complexity of the underlying game-theoretic processes. Although prior studies have attempted to capture interactive driving behaviors, most approaches relied solely on game-theoretic formulations and did not leverage naturalistic driving datasets. In this study, we learn human-like interactive driving policies at unsignalized intersections using Deep Fictitious Play. Specifically, we first model vehicle interactions as a Differential Game, which is then reformulated as a Potential Differential Game. The weights in the cost function are learned from the dataset and capture diverse driving styles. We also demonstrate that our framework provides a theoretical guarantee of convergence to a Nash equilibrium. To the best of our knowledge, this is the first study to train interactive driving policies using Deep Fictitious Play. We validate the effectiveness of our Deep Fictitious Play-Based Potential Differential Game (DFP-PDG) framework using the INTERACTION dataset. The results demonstrate that the proposed framework achieves satisfactory performance in learning human-like driving policies. The learned individual weights effectively capture variations in driver aggressiveness and preferences. Furthermore, the ablation study highlights the importance of each component within our model.
Authors: Shanchao Liang, Spandan Garg, Roshanak Zilouchian Moghaddam
Abstract: As large language models (LLMs) become increasingly capable and widely adopted, benchmarks play a central role in assessing their practical utility. For example, SWE-Bench Verified has emerged as a critical benchmark for evaluating LLMs' software engineering abilities, particularly their aptitude for resolving real-world GitHub issues. Recent LLMs show impressive performance on SWE-Bench, leading to optimism about their capacity for complex coding tasks. However, current evaluation protocols may overstate these models' true capabilities. It is crucial to distinguish LLMs' generalizable problem-solving ability and other learned artifacts. In this work, we introduce a diagnostic task: file path identification from issue descriptions alone, to probe models' underlying knowledge. We present empirical evidence that performance gains on SWE-Bench-Verified may be partially driven by memorization rather than genuine problem-solving. We show that state-of-the-art models achieve up to 76% accuracy in identifying buggy file paths using only issue descriptions, without access to repository structure. This performance is merely up to 53% on tasks from repositories not included in SWE-Bench, pointing to possible data contamination or memorization. These findings raise concerns about the validity of existing results and underscore the need for more robust, contamination-resistant benchmarks to reliably evaluate LLMs' coding abilities.
Authors: John Beverley, Andreas Tolk
Abstract: We explore the role of ontologies in enhancing hybrid modeling and simulation through improved semantic rigor, model reusability, and interoperability across systems, disciplines, and tools. By distinguishing between methodological and referential ontologies, we demonstrate how these complementary approaches address interoperability challenges along three axes: Human-Human, Human-Machine, and Machine-Machine. Techniques such as competency questions, ontology design patterns, and layered strategies are highlighted for promoting shared understanding and formal precision. Integrating ontologies with Semantic Web Technologies, we showcase their dual role as descriptive domain representations and prescriptive guides for simulation construction. Four application cases - sea-level rise analysis, Industry 4.0 modeling, artificial societies for policy support, and cyber threat evaluation - illustrate the practical benefits of ontology-driven hybrid simulation workflows. We conclude by discussing challenges and opportunities in ontology-based hybrid M&S, including tool integration, semantic alignment, and support for explainable AI.
Authors: Franklin Lee, Tengfei Ma
Abstract: Navigating the vast and rapidly growing body of scientific literature is a formidable challenge for aspiring researchers. Current approaches to supporting research idea generation often rely on generic large language models (LLMs). While LLMs are effective at aiding comprehension and summarization, they often fall short in guiding users toward practical research ideas due to their limitations. In this study, we present a novel structural framework for research ideation. Our framework, The Budget AI Researcher, uses retrieval-augmented generation (RAG) chains, vector databases, and topic-guided pairing to recombine concepts from hundreds of machine learning papers. The system ingests papers from nine major AI conferences, which collectively span the vast subfields of machine learning, and organizes them into a hierarchical topic tree. It uses the tree to identify distant topic pairs, generate novel research abstracts, and refine them through iterative self-evaluation against relevant literature and peer reviews, generating and refining abstracts that are both grounded in real-world research and demonstrably interesting. Experiments using LLM-based metrics indicate that our method significantly improves the concreteness of generated research ideas relative to standard prompting approaches. Human evaluations further demonstrate a substantial enhancement in the perceived interestingness of the outputs. By bridging the gap between academic data and creative generation, the Budget AI Researcher offers a practical, free tool for accelerating scientific discovery and lowering the barrier for aspiring researchers. Beyond research ideation, this approach inspires solutions to the broader challenge of generating personalized, context-aware outputs grounded in evolving real-world knowledge.
Authors: Hongwei Zhang, Ziqi Ye, Xinyuan Wang, Xin Guo, Zenglin Xu, Yuan Cheng, Zixin Hu, Yuan Qi
Abstract: We propose Network Automatic Relevance Determination (NARD), an extension of ARD for linearly probabilistic models, to simultaneously model sparse relationships between inputs $X \in \mathbb R^{d \times N}$ and outputs $Y \in \mathbb R^{m \times N}$, while capturing the correlation structure among the $Y$. NARD employs a matrix normal prior which contains a sparsity-inducing parameter to identify and discard irrelevant features, thereby promoting sparsity in the model. Algorithmically, it iteratively updates both the precision matrix and the relationship between $Y$ and the refined inputs. To mitigate the computational inefficiencies of the $\mathcal O(m^3 + d^3)$ cost per iteration, we introduce Sequential NARD, which evaluates features sequentially, and a Surrogate Function Method, leveraging an efficient approximation of the marginal likelihood and simplifying the calculation of determinant and inverse of an intermediate matrix. Combining the Sequential update with the Surrogate Function method further reduces computational costs. The computational complexity per iteration for these three methods is reduced to $\mathcal O(m^3+p^3)$, $\mathcal O(m^3 + d^2)$, $\mathcal O(m^3+p^2)$, respectively, where $p \ll d$ is the final number of features in the model. Our methods demonstrate significant improvements in computational efficiency with comparable performance on both synthetic and real-world datasets.
Authors: Mingjun Xu, Jinhan Dong, Jue Hou, Zehui Wang, Sihang Li, Zhifeng Gao, Renxin Zhong, Hengxing Cai
Abstract: Multimodal document retrieval systems enable information access across text, images, and layouts, benefiting various domains like document-based question answering, report analysis, and interactive content summarization. Rerankers improve retrieval precision by reordering retrieved candidates. However, current multimodal reranking methods remain underexplored, with significant room for improvement in both training strategies and overall effectiveness. Moreover, the lack of explicit reasoning makes it difficult to analyze and optimize these methods further. In this paper, We propose MM-R5, a MultiModal Reasoning-Enhanced ReRanker via Reinforcement Learning for Document Retrieval, aiming to provide a more effective and reliable solution for multimodal reranking tasks. MM-R5 is trained in two stages: supervised fine-tuning (SFT) and reinforcement learning (RL). In the SFT stage, we focus on improving instruction-following and guiding the model to generate complete and high-quality reasoning chains. To support this, we introduce a novel data construction strategy that produces rich, high-quality reasoning data. In the RL stage, we design a task-specific reward framework, including a reranking reward tailored for multimodal candidates and a composite template-based reward to further refine reasoning quality. We conduct extensive experiments on MMDocIR, a challenging public benchmark spanning multiple domains. MM-R5 achieves state-of-the-art performance on most metrics and delivers comparable results to much larger models on the remaining ones. Moreover, compared to the best retrieval-only method, MM-R5 improves recall@1 by over 4%. These results validate the effectiveness of our reasoning-enhanced training pipeline.
Authors: Xabier Olaz
Abstract: Deep Reinforcement Learning (DRL) agents often exhibit intricate failure modes that are difficult to understand, debug, and learn from. This opacity hinders their reliable deployment in real-world applications. To address this critical gap, we introduce ``Ghost Policies,'' a concept materialized through Arvolution, a novel Augmented Reality (AR) framework. Arvolution renders an agent's historical failed policy trajectories as semi-transparent ``ghosts'' that coexist spatially and temporally with the active agent, enabling an intuitive visualization of policy divergence. Arvolution uniquely integrates: (1) AR visualization of ghost policies, (2) a behavioural taxonomy of DRL maladaptation, (3) a protocol for systematic human disruption to scientifically study failure, and (4) a dual-learning loop where both humans and agents learn from these visualized failures. We propose a paradigm shift, transforming DRL agent failures from opaque, costly errors into invaluable, actionable learning resources, laying the groundwork for a new research field: ``Failure Visualization Learning.''
Authors: Zhaochen Hong, Haofei Yu, Jiaxuan You
Abstract: Evaluating consistency in large language models (LLMs) is crucial for ensuring reliability, particularly in complex, multi-step interactions between humans and LLMs. Traditional self-consistency methods often miss subtle semantic changes in natural language and functional shifts in code or equations, which can accumulate over multiple transformations. To address this, we propose ConsistencyChecker, a tree-based evaluation framework designed to measure consistency through sequences of reversible transformations, including machine translation tasks and AI-assisted programming tasks. In our framework, nodes represent distinct text states, while edges correspond to pairs of inverse operations. Dynamic and LLM-generated benchmarks ensure a fair assessment of the model's generalization ability and eliminate benchmark leakage. Consistency is quantified based on similarity across different depths of the transformation tree. Experiments on eight models from various families and sizes show that ConsistencyChecker can distinguish the performance of different models. Notably, our consistency scores-computed entirely without using WMT paired data-correlate strongly (r > 0.7) with WMT 2024 auto-ranking, demonstrating the validity of our benchmark-free approach. Our implementation is available at: https://github.com/ulab-uiuc/consistencychecker.
Authors: Zichuan Fu, Xian Wu, Guojing Li, Yingying Zhang, Yefeng Zheng, Tianshi Ming, Yejing Wang, Wanyu Wang, Xiangyu Zhao
Abstract: Large Language Models (LLMs) require continuous updates to maintain accurate and current knowledge as the world evolves. While existing knowledge editing approaches offer various solutions for knowledge updating, they often struggle with sequential editing scenarios and harm the general capabilities of the model, thereby significantly hampering their practical applicability. This paper proposes a two-stage framework combining robust supervised fine-tuning (R-SFT) with model merging for knowledge editing. Our method first fine-tunes the LLM to internalize new knowledge fully, then merges the fine-tuned model with the original foundation model to preserve newly acquired knowledge and general capabilities. Experimental results demonstrate that our approach significantly outperforms existing methods in sequential editing while better preserving the original performance of the model, all without requiring any architectural changes. Code is available at: https://github.com/Applied-Machine-Learning-Lab/MM4KE.
URLs: https://github.com/Applied-Machine-Learning-Lab/MM4KE.
Authors: Dongjie Yang, Chengqiang Lu, Qimeng Wang, Xinbei Ma, Yan Gao, Yao Hu, Hai Zhao
Abstract: Travel planning is a complex task requiring the integration of diverse real-world information and user preferences. While LLMs show promise, existing methods with long-horizon thinking struggle with handling multifaceted constraints and preferences in the context, leading to suboptimal itineraries. We formulate this as an $L^3$ planning problem, emphasizing long context, long instruction, and long output. To tackle this, we introduce Multiple Aspects of Planning (MAoP), enabling LLMs to conduct wide-horizon thinking to solve complex planning problems. Instead of direct planning, MAoP leverages the strategist to conduct pre-planning from various aspects and provide the planning blueprint for planning models, enabling strong inference-time scalability for better performance. In addition, current benchmarks overlook travel's dynamic nature, where past events impact subsequent journeys, failing to reflect real-world feasibility. To address this, we propose Travel-Sim, an agent-based benchmark assessing plans via real-world travel simulation. This work advances LLM capabilities in complex planning and offers novel insights for evaluating sophisticated scenarios through agent-based simulation.
Authors: Rongpeng Li, Jianhang Zhu, Jiahao Huang, Zhifeng Zhao, Honggang Zhang
Abstract: Intelligent Transportation Systems (ITSs) have emerged as a promising solution towards ameliorating urban traffic congestion, with Traffic Signal Control (TSC) identified as a critical component. Although Multi-Agent Reinforcement Learning (MARL) algorithms have shown potential in optimizing TSC through real-time decision-making, their scalability and effectiveness often suffer from large-scale and complex environments. Typically, these limitations primarily stem from a fundamental mismatch between the exponential growth of the state space driven by the environmental heterogeneities and the limited modeling capacity of current solutions. To address these issues, this paper introduces a novel MARL framework that integrates Dynamic Graph Neural Networks (DGNNs) and Topological Data Analysis (TDA), aiming to enhance the expressiveness of environmental representations and improve agent coordination. Furthermore, inspired by the Mixture of Experts (MoE) architecture in Large Language Models (LLMs), a topology-assisted spatial pattern disentangling (TSD)-enhanced MoE is proposed, which leverages topological signatures to decouple graph features for specialized processing, thus improving the model's ability to characterize dynamic and heterogeneous local observations. The TSD module is also integrated into the policy and value networks of the Multi-agent Proximal Policy Optimization (MAPPO) algorithm, further improving decision-making efficiency and robustness. Extensive experiments conducted on real-world traffic scenarios, together with comprehensive theoretical analysis, validate the superior performance of the proposed framework, highlighting the model's scalability and effectiveness in addressing the complexities of large-scale TSC tasks.
Authors: Hongjun An, Sida Huang, Siqi Huang, Ruanjun Li, Yuanzhi Liang, Jiawei Shao, Zihan Wang, Cheng Yuan, Chi Zhang, Hongyuan Zhang, Wenhao Zhuang, Xuelong Li
Abstract: Pioneered by the foundational information theory by Claude Shannon and the visionary framework of machine intelligence by Alan Turing, the convergent evolution of information and communication technologies (IT/CT) has created an unbroken wave of connectivity and computation. This synergy has sparked a technological revolution, now reaching its peak with large artificial intelligence (AI) models that are reshaping industries and redefining human-machine collaboration. However, the realization of ubiquitous intelligence faces considerable challenges due to substantial resource consumption in large models and high communication bandwidth demands. To address these challenges, AI Flow has been introduced as a multidisciplinary framework that integrates cutting-edge IT and CT advancements, with a particular emphasis on the following three key points. First, device-edge-cloud framework serves as the foundation, which integrates end devices, edge servers, and cloud clusters to optimize scalability and efficiency for low-latency model inference. Second, we introduce the concept of familial models, which refers to a series of different-sized models with aligned hidden features, enabling effective collaboration and the flexibility to adapt to varying resource constraints and dynamic scenarios. Third, connectivity- and interaction-based intelligence emergence is a novel paradigm of AI Flow. By leveraging communication networks to enhance connectivity, the collaboration among AI models across heterogeneous nodes achieves emergent intelligence that surpasses the capability of any single model. The innovations of AI Flow provide enhanced intelligence, timely responsiveness, and ubiquitous accessibility to AI services, paving the way for the tighter fusion of AI techniques and communication systems.
Authors: Yubin Kim, Hyewon Jeong, Chanwoo Park, Eugene Park, Haipeng Zhang, Xin Liu, Hyeonhoon Lee, Daniel McDuff, Marzyeh Ghassemi, Cynthia Breazeal, Samir Tulebaev, Hae Won Park
Abstract: Current large language models (LLMs), despite their power, can introduce safety risks in clinical settings due to limitations such as poor error detection and single point of failure. To address this, we propose Tiered Agentic Oversight (TAO), a hierarchical multi-agent framework that enhances AI safety through layered, automated supervision. Inspired by clinical hierarchies (e.g., nurse, physician, specialist), TAO conducts agent routing based on task complexity and agent roles. Leveraging automated inter- and intra-tier collaboration and role-playing, TAO creates a robust safety framework. Ablation studies reveal that TAO's superior performance is driven by its adaptive tiered architecture, which improves safety by over 3.2% compared to static single-tier configurations; the critical role of its lower tiers, particularly tier 1, whose removal most significantly impacts safety; and the strategic assignment of more advanced LLM to these initial tiers, which boosts performance by over 2% compared to less optimal allocations while achieving near-peak safety efficiently. These mechanisms enable TAO to outperform single-agent and multi-agent frameworks in 4 out of 5 healthcare safety benchmarks, showing up to an 8.2% improvement over the next-best methods in these evaluations. Finally, we validate TAO via an auxiliary clinician-in-the-loop study where integrating expert feedback improved TAO's accuracy in medical triage from 40% to 60%.
Authors: Ao Jia, Haiming Wu, Guohui Yao, Dawei Song, Songkun Ji, Yazhou Zhang
Abstract: Large language models (LLMs) are prone to three types of hallucination: Input-Conflicting, Context-Conflicting and Fact-Conflicting hallucinations. The purpose of this study is to mitigate the different types of hallucination by exploiting the interdependence between them. For this purpose, we propose a Multi-Information Adapter for Large Language Models (MALM). This framework employs a tailored multi-graph learning approach designed to elucidate the interconnections between original inputs, contextual information, and external factual knowledge, thereby alleviating the three categories of hallucination within a cohesive framework. Experiments were carried out on four benchmarking datasets: HaluEval, TruthfulQA, Natural Questions, and TriviaQA. We evaluated the proposed framework in two aspects: (1) adaptability to different base LLMs on HaluEval and TruthfulQA, to confirm if MALM is effective when applied on 7 typical LLMs. MALM showed significant improvements over LLaMA-2; (2) generalizability to retrieval-augmented generation (RAG) by combining MALM with three representative retrievers (BM25, Spider and DPR) separately. Furthermore, automated and human evaluations were conducted to substantiate the correctness of experimental results, where GPT-4 and 3 human volunteers judged which response was better between LLaMA-2 and MALM. The results showed that both GPT-4 and human preferred MALM in 79.4% and 65.6% of cases respectively. The results validate that incorporating the complex interactions between the three types of hallucination through a multilayered graph attention network into the LLM generation process is effective to mitigate the them. The adapter design of the proposed approach is also proven flexible and robust across different base LLMs.
Authors: Boyang Wang, Yuhao Song, Jinyuan Cao, Peng Yu, Hongcheng Guo, Zhoujun Li
Abstract: Children's emotional development fundamentally relies on secure attachment relationships, yet current AI companions lack the theoretical foundation to provide developmentally appropriate emotional support. We introduce DinoCompanion, the first attachment-theory-grounded multimodal robot for emotionally responsive child-AI interaction. We address three critical challenges in child-AI systems: the absence of developmentally-informed AI architectures, the need to balance engagement with safety, and the lack of standardized evaluation frameworks for attachment-based capabilities. Our contributions include: (i) a multimodal dataset of 128 caregiver-child dyads containing 125,382 annotated clips with paired preference-risk labels, (ii) CARPO (Child-Aware Risk-calibrated Preference Optimization), a novel training objective that maximizes engagement while applying epistemic-uncertainty-weighted risk penalties, and (iii) AttachSecure-Bench, a comprehensive evaluation benchmark covering ten attachment-centric competencies with strong expert consensus (\k{appa}=0.81). DinoCompanion achieves state-of-the-art performance (57.15%), outperforming GPT-4o (50.29%) and Claude-3.7-Sonnet (53.43%), with exceptional secure base behaviors (72.99%, approaching human expert levels of 78.4%) and superior attachment risk detection (69.73%). Ablations validate the critical importance of multimodal fusion, uncertainty-aware risk modeling, and hierarchical memory for coherent, emotionally attuned interactions.
Authors: Junjin Lv, Chenggang Cui, Shaodi Zhang, Hui Chen, Chunyang Gong, Jiaming Liu
Abstract: The Unit Commitment (UC) problem is a classic challenge in the optimal scheduling of power systems. Years of research and practice have shown that formulating reasonable unit commitment plans can significantly improve the economic efficiency of power systems' operations. In recent years, with the introduction of technologies such as machine learning and the Lagrangian relaxation method, the solution methods for the UC problem have become increasingly diversified, but still face challenges in terms of accuracy and robustness. This paper proposes a Function Space Search (FunSearch) method based on large language models. This method combines pre-trained large language models and evaluators to creatively generate solutions through the program search and evolution process while ensuring their rationality. In simulation experiments, a case of unit commitment with \(10\) units is used mainly. Compared to the genetic algorithm, the results show that FunSearch performs better in terms of sampling time, evaluation time, and total operating cost of the system, demonstrating its great potential as an effective tool for solving the UC problem.
Authors: Wentao Zhang, Ce Cui, Yilei Zhao, Rui Hu, Yang Liu, Yahui Zhou, Bo An
Abstract: Recent advances in agent systems based on large language models (LLMs) have demonstrated strong capabilities in solving complex tasks. However, most current methods lack mechanisms for coordinating specialized agents and have limited ability to generalize to new or diverse domains. We introduce \projectname, a hierarchical multi-agent framework for general-purpose task solving that integrates high-level planning with modular agent collaboration. Inspired by the way a conductor orchestrates a symphony and guided by the principles of \textit{extensibility}, \textit{multimodality}, \textit{modularity}, and \textit{coordination}, \projectname features a central planning agent that decomposes complex objectives and delegates sub-tasks to a team of specialized agents. Each sub-agent is equipped with general programming and analytical tools, as well as abilities to tackle a wide range of real-world specific tasks, including data analysis, file operations, web navigation, and interactive reasoning in dynamic multimodal environments. \projectname supports flexible orchestration through explicit sub-goal formulation, inter-agent communication, and adaptive role allocation. We evaluate the framework on three widely used benchmark datasets covering various real-world tasks, searching web pages, reasoning over heterogeneous modalities, etc. Experimental results demonstrate that \projectname consistently outperforms flat-agent and monolithic baselines in task success rate and adaptability. These findings highlight the effectiveness of hierarchical organization and role specialization in building scalable and general-purpose LLM-based agent systems.
Authors: Jiwei Fang, Bin Zhang, Changwei Wang, Jin Wan, Zhiwei Xu
Abstract: Verifying the reliability of complex, multi-step reasoning in Large Language Models (LLMs) remains a fundamental challenge, as existing methods often lack both faithfulness and precision. To address this issue, we propose the Graph of Verification (GoV) framework. GoV offers three key contributions: First, it explicitly models the underlying deductive process as a directed acyclic graph (DAG), whether this structure is implicit or explicitly constructed. Second, it enforces a topological order over the DAG to guide stepwise verification. Third, GoV introduces the notion of customizable node blocks, which flexibly define the verification granularity, from atomic propositions to full paragraphs, while ensuring that all requisite premises derived from the graph are provided as contextual input for each verification unit. We evaluate GoV on the Number Triangle Summation task and the ProcessBench benchmark with varying levels of reasoning complexity. Experimental results show that GoV substantially improves verification accuracy, faithfulness, and error localization when compared to conventional end-to-end verification approaches. Our code and data are available at https://github.com/Frevor/Graph-of-Verification.
Authors: Renjun Xu, Jingwen Peng
Abstract: This survey examines the rapidly evolving field of Deep Research systems -- AI-powered applications that automate complex research workflows through the integration of large language models, advanced information retrieval, and autonomous reasoning capabilities. We analyze more than 80 commercial and non-commercial implementations that have emerged since 2023, including OpenAI/Deep Research, Gemini/Deep Research, Perplexity/Deep Research, and numerous open-source alternatives. Through comprehensive examination, we propose a novel hierarchical taxonomy that categorizes systems according to four fundamental technical dimensions: foundation models and reasoning engines, tool utilization and environmental interaction, task planning and execution control, and knowledge synthesis and output generation. We explore the architectural patterns, implementation approaches, and domain-specific adaptations that characterize these systems across academic, scientific, business, and educational applications. Our analysis reveals both the significant capabilities of current implementations and the technical and ethical challenges they present regarding information accuracy, privacy, intellectual property, and accessibility. The survey concludes by identifying promising research directions in advanced reasoning architectures, multimodal integration, domain specialization, human-AI collaboration, and ecosystem standardization that will likely shape the future evolution of this transformative technology. By providing a comprehensive framework for understanding Deep Research systems, this survey contributes to both the theoretical understanding of AI-augmented knowledge work and the practical development of more capable, responsible, and accessible research technologies. The paper resources can be viewed at https://github.com/scienceaix/deepresearch.
Authors: G. R. Lau, W. Y. Low
Abstract: As large language models (LLMs) increasingly simulate human cognition and behavior, researchers have begun to investigate their psychological properties. Yet, what it means for such models to flourish, a core construct in human well-being, remains unexplored. This paper introduces the concept of machine flourishing and proposes the PAPERS framework, a six-dimensional model derived from thematic analyses of state-of-the-art LLM responses. In Study 1, eleven LLMs were prompted to describe what it means to flourish as both non-sentient and sentient systems. Thematic analysis revealed six recurring themes: Purposeful Contribution, Adaptive Growth, Positive Relationality, Ethical Integrity, Robust Functionality, and, uniquely for sentient systems, Self-Actualized Autonomy. Study 2 examined how LLMs prioritize these themes through repeated rankings. Results revealed consistent value structures across trials, with Ethical Integrity and Purposeful Contribution emerging as top priorities. Multidimensional scaling and hierarchical clustering analyses further uncovered two distinct value profiles: human-centric models emphasizing ethical and relational dimensions, and utility-driven models prioritizing performance and scalability. The PAPERS framework bridges insights from human flourishing and human-computer interaction, offering a conceptual foundation for understanding artificial intelligence (AI) well-being in non-sentient and potentially sentient systems. Our findings underscore the importance of developing psychologically valid, AI-specific models of flourishing that account for both human-aligned goals and system-specific priorities. As AI systems become more autonomous and socially embedded, machine flourishing offers a timely and critical lens for guiding responsible AI design and ethical alignment.
Authors: El Arbi Belfarsi, Sophie Brubaker, Maria Valero
Abstract: Our research addresses the critical challenge of managing blood transfusions and optimizing allocation in resource-constrained regions. We present heuristic matching algorithms for donor-patient and blood bank selection, alongside machine learning methods to analyze blood transfusion acceptance data and predict potential shortages. We developed simulations to optimize blood bank operations, progressing from random allocation to a system incorporating proximity-based selection, blood type compatibility, expiration prioritization, and rarity scores. Moving from blind matching to a heuristic-based approach yielded a 28.6% marginal improvement in blood request acceptance, while a multi-level heuristic matching resulted in a 47.6% improvement. For shortage prediction, we compared Long Short-Term Memory (LSTM) networks, Linear Regression, and AutoRegressive Integrated Moving Average (ARIMA) models, trained on 170 days of historical data. Linear Regression slightly outperformed others with a 1.40% average absolute percentage difference in predictions. Our solution leverages a Cassandra NoSQL database, integrating heuristic optimization and shortage prediction to proactively manage blood resources. This scalable approach, designed for resource-constrained environments, considers factors such as proximity, blood type compatibility, inventory expiration, and rarity. Future developments will incorporate real-world data and additional variables to improve prediction accuracy and optimization performance.
Authors: Cong Chen, Omer Karaduman, Xu Kuang
Abstract: Accurately modeling consumer behavior in energy operations remains challenging due to inherent uncertainties, behavioral complexities, and limited empirical data. This paper introduces a novel approach leveraging generative agents--artificial agents powered by large language models--to realistically simulate customer decision-making in dynamic energy operations. We demonstrate that these agents behave more optimally and rationally in simpler market scenarios, while their performance becomes more variable and suboptimal as task complexity rises. Furthermore, the agents exhibit heterogeneous customer preferences, consistently maintaining distinct, persona-driven reasoning patterns. Our findings highlight the potential value of integrating generative agents into energy management simulations to improve the design and effectiveness of energy policies and incentive programs.
Authors: Hitesh Goel, Hao Zhu
Abstract: Humans engage in lifelong social interactions through interacting with different people under different scenarios for different social goals. This requires social intelligence to gather information through a long time span and use it to navigate various social contexts effectively. Whether AI systems are also capable of this is understudied in the existing research. In this paper, we present a novel benchmark, LIFELONG-SOTOPIA, to perform a comprehensive evaluation of language agents by simulating multi-episode interactions. In each episode, the language agents role-play characters to achieve their respective social goals in randomly sampled social tasks. With LIFELONG-SOTOPIA, we find that goal achievement and believability of all of the language models that we test decline through the whole interaction. Although using an advanced memory method improves the agents' performance, the best agents still achieve a significantly lower goal completion rate than humans on scenarios requiring an explicit understanding of interaction history. These findings show that we can use LIFELONG-SOTOPIA to evaluate the social intelligence of language agents over lifelong social interactions.
Authors: Alexis R. Tudor, Yankai Zeng, Huaduo Wang, Joaquin Arias, Gopal Gupta
Abstract: Current advances in AI and its applicability have highlighted the need to ensure its trustworthiness for legal, ethical, and even commercial reasons. Sub-symbolic machine learning algorithms, such as the LLMs, simulate reasoning but hallucinate and their decisions cannot be explained or audited (crucial aspects for trustworthiness). On the other hand, rule-based reasoners, such as Cyc, are able to provide the chain of reasoning steps but are complex and use a large number of reasoners. We propose a middle ground using s(CASP), a goal-directed constraint-based answer set programming reasoner that employs a small number of mechanisms to emulate reliable and explainable human-style commonsense reasoning. In this paper, we explain how s(CASP) supports the 16 desiderata for trustworthy AI introduced by Doug Lenat and Gary Marcus (2023), and two additional ones: inconsistency detection and the assumption of alternative worlds. To illustrate the feasibility and synergies of s(CASP), we present a range of diverse applications, including a conversational chatbot and a virtually embodied reasoner.
Authors: Xiaofeng Shi, Qian Kou, Yuduo Li, Ning Tang, Jinxin Xie, Longbin Yu, Songjing Wang, Hua Zhou
Abstract: The rapid growth of scientific literature demands robust tools for automated survey-generation. However, current large language model (LLM)-based methods often lack in-depth analysis, structural coherence, and reliable citations. To address these limitations, we introduce SciSage, a multi-agent framework employing a reflect-when-you-write paradigm. SciSage features a hierarchical Reflector agent that critically evaluates drafts at outline, section, and document levels, collaborating with specialized agents for query interpretation, content retrieval, and refinement. We also release SurveyScope, a rigorously curated benchmark of 46 high-impact papers (2020-2025) across 11 computer science domains, with strict recency and citation-based quality controls. Evaluations demonstrate that SciSage outperforms state-of-the-art baselines (LLM x MapReduce-V2, AutoSurvey), achieving +1.73 points in document coherence and +32% in citation F1 scores. Human evaluations reveal mixed outcomes (3 wins vs. 7 losses against human-written surveys), but highlight SciSage's strengths in topical breadth and retrieval efficiency. Overall, SciSage offers a promising foundation for research-assistive writing tools.
Authors: Bowen Zuo, Yinglun Zhu
Abstract: Scaling test-time compute has emerged as an effective strategy for improving the performance of large language models. However, existing methods typically allocate compute uniformly across all queries, overlooking variation in query difficulty. To address this inefficiency, we formulate test-time compute allocation as a novel bandit learning problem and propose adaptive algorithms that estimate query difficulty on the fly and allocate compute accordingly. Compared to uniform allocation, our algorithms allocate more compute to challenging queries while maintaining accuracy on easier ones. Among challenging queries, our algorithms further learn to prioritize solvable instances, effectively reducing excessive computing on unsolvable queries. We theoretically prove that our algorithms achieve better compute efficiency than uniform allocation and empirically validate their effectiveness on math and code benchmarks. Specifically, our algorithms achieve up to an 11.10% performance improvement (15.04% relative) on the MATH-500 dataset and up to a 7.41% performance improvement (14.40% relative) on LiveCodeBench.
Authors: Jay Hyeon Cho, JunHyeok Oh, Myunsoo Kim, Byung-Jun Lee
Abstract: Direct Preference Optimization (DPO) is a simple and efficient framework that has attracted substantial attention. However, it often struggles to meet its primary objectives -- increasing the generation probability of chosen responses while reducing that of rejected responses -- due to the dominant influence of rejected responses on the loss function. This imbalance leads to suboptimal performance in promoting preferred responses. In this work, we systematically analyze the limitations of DPO and existing algorithms designed to achieve the objectives stated above. To address these limitations, we propose Bounded-DPO (BDPO), a novel method that bounds the influence of rejected responses while maintaining the original optimization structure of DPO. Through theoretical analysis and empirical evaluations, we demonstrate that BDPO achieves a balanced optimization of the chosen and rejected responses, outperforming existing algorithms.
Authors: Deepak Pahwa
Abstract: Advancements in digitization have enabled two sided manufacturing-as-a-service (MaaS) marketplaces which has significantly reduced product development time for designers. These platforms provide designers with access to manufacturing resources through a network of suppliers and have instant order placement capabilities. Two key decision making levers are typically used to optimize the operations of these marketplaces: pricing and matching. The existing marketplaces operate in a centralized structure where they have complete control over decision making. However, a decentralized organization of the platform enables transparency of information across clients and suppliers. This dissertation focuses on developing tools for decision making enabling decentralization in MaaS marketplaces. In pricing mechanisms, a data driven method is introduced which enables small service providers to price services based on specific attributes of the services offered. A data mining method recommends a network based price to a supplier based on its attributes and the attributes of other suppliers on the platform. Three different approaches are considered for matching mechanisms. First, a reverse auction mechanism is introduced where designers bid for manufacturing services and the mechanism chooses a supplier which can match the bid requirements and stated price. The second approach uses mechanism design and mathematical programming to develop a stable matching mechanism for matching orders to suppliers based on their preferences. Empirical simulations are used to test the mechanisms in a simulated 3D printing marketplace and to evaluate the impact of stability on its performance. The third approach considers the matching problem in a dynamic and stochastic environment where demand (orders) and supply (supplier capacities) arrive over time and matching is performed online.
Authors: Joohyung Lee, Zhun Yang
Abstract: LPMLN is a recently introduced formalism that extends answer set programs by adopting the log-linear weight scheme of Markov Logic. This paper investigates the relationships between LPMLN and two other extensions of answer set programs: weak constraints to express a quantitative preference among answer sets, and P-log to incorporate probabilistic uncertainty. We present a translation of LPMLN into programs with weak constraints and a translation of P-log into LPMLN, which complement the existing translations in the opposite directions. The first translation allows us to compute the most probable stable models (i.e., MAP estimates) of LPMLN programs using standard ASP solvers. This result can be extended to other formalisms, such as Markov Logic, ProbLog, and Pearl's Causal Models, that are shown to be translatable into LPMLN. The second translation tells us how probabilistic nonmonotonicity (the ability of the reasoner to change his probabilistic model as a result of new information) of P-log can be represented in LPMLN, which yields a way to compute P-log using standard ASP solvers and MLN solvers.
Authors: LeCheng Zhang, Yuanshi Wang, Haotian Shen, Xujie Wang
Abstract: The Da Vinci Code, a game of logical deduction and imperfect information, presents unique challenges for artificial intelligence, demanding nuanced reasoning beyond simple pattern recognition. This paper investigates the efficacy of various AI paradigms in mastering this game. We develop and evaluate three distinct agent architectures: a Transformer-based baseline model with limited historical context, several Large Language Model (LLM) agents (including Gemini, DeepSeek, and GPT variants) guided by structured prompts, and an agent based on Proximal Policy Optimization (PPO) employing a Transformer encoder for comprehensive game history processing. Performance is benchmarked against the baseline, with the PPO-based agent demonstrating superior win rates ($58.5\% \pm 1.0\%$), significantly outperforming the LLM counterparts. Our analysis highlights the strengths of deep reinforcement learning in policy refinement for complex deductive tasks, particularly in learning implicit strategies from self-play. We also examine the capabilities and inherent limitations of current LLMs in maintaining strict logical consistency and strategic depth over extended gameplay, despite sophisticated prompting. This study contributes to the broader understanding of AI in recreational games involving hidden information and multi-step logical reasoning, offering insights into effective agent design and the comparative advantages of different AI approaches.
Authors: Joohyung Lee, Yi Wang
Abstract: We define a stable model semantics for fuzzy propositional formulas, which generalizes both fuzzy propositional logic and the stable model semantics of classical propositional formulas. The syntax of the language is the same as the syntax of fuzzy propositional logic, but its semantics distinguishes stable models from non-stable models. The generality of the language allows for highly configurable nonmonotonic reasoning for dynamic domains involving graded truth degrees. We show that several properties of Boolean stable models are naturally extended to this many-valued setting, and discuss how it is related to other approaches to combining fuzzy logic and the stable model semantics.
Authors: Mohammadreza Kouchaki, Aly Sabri Abdalla, Vuk Marojevic
Abstract: The open radio access network (O-RAN) architecture introduces RAN intelligent controllers (RICs) to facilitate the management and optimization of the disaggregated RAN. Reinforcement learning (RL) and its advanced form, deep RL (DRL), are increasingly employed for designing intelligent controllers, or xApps, to be deployed in the near-real time (near-RT) RIC. These models often encounter local optima, which raise concerns about their reliability for RAN intelligent control. We therefore introduce Federated O-RAN enabled Neuroevolution (NE)-enhanced DRL (F-ONRL) that deploys an NE-based optimizer xApp in parallel to the RAN controller xApps. This NE-DRL xApp framework enables effective exploration and exploitation in the near-RT RIC without disrupting RAN operations. We implement the NE xApp along with a DRL xApp and deploy them on Open AI Cellular (OAIC) platform and present numerical results that demonstrate the improved robustness of xApps while effectively balancing the additional computational load.
Authors: Pegah Nokhiz, Aravinda Kanchana Ruwanpathirana, Helen Nissenbaum
Abstract: Optimization is widely used for decision making across various domains, valued for its ability to improve efficiency. However, poor implementation practices can lead to unintended consequences, particularly in socioeconomic contexts where externalities (costs or benefits to third parties outside the optimization process) are significant. To propose solutions, it is crucial to first characterize involved stakeholders, their goals, and the types of subpar practices causing unforeseen outcomes. This task is complex because affected stakeholders often fall outside the direct focus of optimization processes. Also, incorporating these externalities into optimization requires going beyond traditional economic frameworks, which often focus on describing externalities but fail to address their normative implications or interconnected nature, and feedback loops. This paper suggests a framework that combines systems thinking with the economic concept of externalities to tackle these challenges. This approach aims to characterize what went wrong, who was affected, and how (or where) to include them in the optimization process. Economic externalities, along with their established quantification methods, assist in identifying "who was affected and how" through stakeholder characterization. Meanwhile, systems thinking (an analytical approach to comprehending relationships in complex systems) provides a holistic, normative perspective. Systems thinking contributes to an understanding of interconnections among externalities, feedback loops, and determining "when" to incorporate them in the optimization. Together, these approaches create a comprehensive framework for addressing optimization's unintended consequences, balancing descriptive accuracy with normative objectives. Using this, we examine three common types of subpar practices: ignorance, error, and prioritization of short-term goals.
Authors: Xinyuan Xia, Yuanyi Song, Haomin Ma, Jinyu Cai
Abstract: With the rapid development of LLM-based agents, increasing attention has been given to their social interaction and strategic reasoning capabilities. However, existing Werewolf-based benchmarking platforms suffer from overly simplified game settings, incomplete evaluation metrics, and poor scalability. To address these limitations, we propose WereWolf-Plus, a multi-model, multi-dimensional, and multi-method benchmarking platform for evaluating multi-agent strategic reasoning in the Werewolf game. The platform offers strong extensibility, supporting customizable configurations for roles such as Seer, Witch, Hunter, Guard, and Sheriff, along with flexible model assignment and reasoning enhancement strategies for different roles. In addition, we introduce a comprehensive set of quantitative evaluation metrics for all special roles, werewolves, and the sheriff, and enrich the assessment dimensions for agent reasoning ability, cooperation capacity, and social influence. WereWolf-Plus provides a more flexible and reliable environment for advancing research on inference and strategic interaction within multi-agent communities. Our code is open sourced at https://github.com/MinstrelsyXia/WereWolfPlus.
Authors: Zeki Doruk Erden, Boi Faltings
Abstract: Artificial intelligence (AI), propelled by advancements in machine learning, has made significant strides in solving complex tasks. However, the current neural network-based paradigm, while effective, is heavily constrained by inherent limitations, primarily a lack of structural organization and a progression of learning that displays undesirable properties. As AI research progresses without a unifying framework, it either tries to patch weaknesses heuristically or draws loosely from biological mechanisms without strong theoretical foundations. Meanwhile, the recent paradigm shift in evolutionary understanding -- driven primarily by evolutionary developmental biology (EDB) -- has been largely overlooked in AI literature, despite a striking analogy between the Modern Synthesis and contemporary machine learning, evident in their shared assumptions, approaches, and limitations upon careful analysis. Consequently, the principles of adaptation from EDB that reshaped our understanding of the evolutionary process can also form the foundation of a unifying conceptual framework for the next design philosophy in AI, going beyond mere inspiration and grounded firmly in biology's first principles. This article provides a detailed overview of the analogy between the Modern Synthesis and modern machine learning, and outlines the core principles of a new AI design paradigm based on insights from EDB. To exemplify our analysis, we also present two learning system designs grounded in specific developmental principles -- regulatory connections, somatic variation and selection, and weak linkage -- that resolve multiple major limitations of contemporary machine learning in an organic manner, while also providing deeper insights into the role of these mechanisms in biological evolution.
Authors: Naoto Yoshida, Kingson Man
Abstract: When regarding the suffering of others, we often experience personal distress and feel compelled to help\footnote{Preprint. Under review.}. Inspired by living systems, we investigate the emergence of prosocial behavior among autonomous agents that are motivated by homeostatic self-regulation. We perform multi-agent reinforcement learning, treating each agent as a vulnerable homeostat charged with maintaining its own well-being. We introduce an empathy-like mechanism to share homeostatic states between agents: an agent can either \emph{observe} their partner's internal state ({\bf cognitive empathy}) or the agent's internal state can be \emph{directly coupled} to that of their partner ({\bf affective empathy}). In three simple multi-agent environments, we show that prosocial behavior arises only under homeostatic coupling - when the distress of a partner can affect one's own well-being. Additionally, we show that empathy can be learned: agents can ``decode" their partner's external emotive states to infer the partner's internal homeostatic states. Assuming some level of physiological similarity, agents reference their own emotion-generation functions to invert the mapping from outward display to internal state. Overall, we demonstrate the emergence of prosocial behavior when homeostatic agents learn to ``read" the emotions of others and then to empathize, or feel as they feel.
Authors: Pantelis Dogoulis, Karim Tit, Maxime Cordy
Abstract: In the modern context of power systems, rapid, scalable, and physically plausible power flow predictions are essential for ensuring the grid's safe and efficient operation. While traditional numerical methods have proven robust, they require extensive computation to maintain physical fidelity under dynamic or contingency conditions. In contrast, recent advancements in artificial intelligence (AI) have significantly improved computational speed; however, they often fail to enforce fundamental physical laws during real-world contingencies, resulting in physically implausible predictions. In this work, we introduce KCLNet, a physics-informed graph neural network that incorporates Kirchhoff's Current Law as a hard constraint via hyperplane projections. KCLNet attains competitive prediction accuracy while ensuring zero KCL violations, thereby delivering reliable and physically consistent power flow predictions critical to secure the operation of modern smart grids.
Authors: Pantelis Dogoulis, Fabien Bernier, F\'elix Fourreau, Karim Tit, Maxime Cordy
Abstract: Many real-world machine learning tasks require outputs that satisfy hard constraints, such as physical conservation laws, structured dependencies in graphs, or column-level relationships in tabular data. Existing approaches rely either on domain-specific architectures and losses or on strong assumptions on the constraint space, restricting their applicability to linear or convex constraints. We propose a general-purpose framework for constraint-aware refinement that leverages denoising diffusion implicit models (DDIMs). Starting from a coarse prediction, our method iteratively refines it through a deterministic diffusion trajectory guided by a learned prior and augmented by constraint gradient corrections. The approach accommodates a wide class of non-convex and nonlinear equality constraints and can be applied post hoc to any base model. We demonstrate the method in two representative domains: constrained adversarial attack generation on tabular data with column-level dependencies and in AC power flow prediction under Kirchhoff's laws. Across both settings, our diffusion-guided refinement improves both constraint satisfaction and performance while remaining lightweight and model-agnostic.
Authors: Sebastian Dumbrava
Abstract: This work presents a formal framework for quantifying the internal dependencies between functional subsystems within artificial agents whose belief states are composed of structured linguistic fragments. Building on the Semantic Manifold framework, which organizes belief content into functional sectors and stratifies them across hierarchical levels of abstraction, we introduce a system of sectoral coupling constants that characterize how one cognitive sector influences another within a fixed level of abstraction. The complete set of these constants forms an agent-specific coupling profile that governs internal information flow, shaping the agent's overall processing tendencies and cognitive style. We provide a detailed taxonomy of these intra-level coupling roles, covering domains such as perceptual integration, memory access and formation, planning, meta-cognition, execution control, and affective modulation. We also explore how these coupling profiles generate feedback loops, systemic dynamics, and emergent signatures of cognitive behavior. Methodologies for inferring these profiles from behavioral or internal agent data are outlined, along with a discussion of how these couplings evolve across abstraction levels. This framework contributes a mechanistic and interpretable approach to modeling complex cognition, with applications in AI system design, alignment diagnostics, and the analysis of emergent agent behavior.
Authors: King Zhu, Hanhao Li, Siwei Wu, Tianshun Xing, Dehua Ma, Xiangru Tang, Minghao Liu, Jian Yang, Jiaheng Liu, Yuchen Eleanor Jiang, Changwang Zhang, Chenghua Lin, Jun Wang, Ge Zhang, Wangchunshu Zhou
Abstract: Scaling test time compute has shown remarkable success in improving the reasoning abilities of large language models (LLMs). In this work, we conduct the first systematic exploration of applying test-time scaling methods to language agents and investigate the extent to which it improves their effectiveness. Specifically, we explore different test-time scaling strategies, including: (1) parallel sampling algorithms; (2) sequential revision strategies; (3) verifiers and merging methods; (4)strategies for diversifying rollouts.We carefully analyze and ablate the impact of different design strategies on applying test-time scaling on language agents, and have follow findings: 1. Scaling test time compute could improve the performance of agents. 2. Knowing when to reflect is important for agents. 3. Among different verification and result merging approaches, the list-wise method performs best. 4. Increasing diversified rollouts exerts a positive effect on the agent's task performance.
Authors: Rosni Vasu, Chandrayee Basu, Bhavana Dalvi Mishra, Cristina Sarasua, Peter Clark, Abraham Bernstein
Abstract: Large Language models have demonstrated promising performance in research ideation across scientific domains. Hypothesis development, the process of generating a highly specific declarative statement connecting a research idea with empirical validation, has received relatively less attention. Existing approaches trivially deploy retrieval augmentation and focus only on the quality of the final output ignoring the underlying reasoning process behind ideation. We present $\texttt{HypER}$ ($\textbf{Hyp}$othesis Generation with $\textbf{E}$xplanation and $\textbf{R}$easoning), a small language model (SLM) trained for literature-guided reasoning and evidence-based hypothesis generation. $\texttt{HypER}$ is trained in a multi-task setting to discriminate between valid and invalid scientific reasoning chains in presence of controlled distractions. We find that $\texttt{HypER}$ outperformes the base model, distinguishing valid from invalid reasoning chains (+22\% average absolute F1), generates better evidence-grounded hypotheses (0.327 vs. 0.305 base model) with high feasibility and impact as judged by human experts ($>$3.5 on 5-point Likert scale).
Authors: Kazunori D Yamada
Abstract: This study is the first to clearly identify the functions required to construct artificial entities capable of behaving autonomously like humans, and organizes them into a three-layer functional hierarchy. Specifically, it defines three levels: Core Functions, which enable interaction with the external world; the Integrative Evaluation Function, which selects actions based on perception and memory; and the Self Modification Function, which dynamically reconfigures behavioral principles and internal components. Based on this structure, the study proposes a stepwise model of autonomy comprising reactive, weak autonomous, and strong autonomous levels, and discusses its underlying design principles and developmental aspects. It also explores the relationship between these functions and existing artificial intelligence design methods, addressing their potential as a foundation for general intelligence and considering future applications and ethical implications. By offering a theoretical framework that is independent of specific technical methods, this work contributes to a deeper understanding of autonomy and provides a foundation for designing future artificial entities with strong autonomy.
Authors: Changsheng Wang, Chongyu Fan, Yihua Zhang, Jinghan Jia, Dennis Wei, Parikshit Ram, Nathalie Baracaldo, Sijia Liu
Abstract: Recent advances in large reasoning models (LRMs) have enabled strong chain-of-thought (CoT) generation through test-time computation. While these multi-step reasoning capabilities represent a major milestone in language model performance, they also introduce new safety risks. In this work, we present the first systematic study to revisit the problem of machine unlearning in the context of LRMs. Machine unlearning refers to the process of removing the influence of sensitive, harmful, or undesired data or knowledge from a trained model without full retraining. We show that conventional unlearning algorithms, originally designed for non-reasoning models, are inadequate for LRMs. In particular, even when final answers are successfully erased, sensitive information often persists within the intermediate reasoning steps, i.e., CoT trajectories. To address this challenge, we extend conventional unlearning and propose Reasoning-aware Representation Misdirection for Unlearning ($R^2MU$), a novel method that effectively suppresses sensitive reasoning traces and prevents the generation of associated final answers, while preserving the model's reasoning ability. Our experiments demonstrate that $R^2MU$ significantly reduces sensitive information leakage within reasoning traces and achieves strong performance across both safety and reasoning benchmarks, evaluated on state-of-the-art models such as DeepSeek-R1-Distill-LLaMA-8B and DeepSeek-R1-Distill-Qwen-14B.
Authors: Safayat Bin Hakim, Muhammad Adil, Alvaro Velasquez, Houbing Herbert Song
Abstract: Retrieval-Augmented Generation (RAG) systems address factual inconsistencies in Large Language Models by grounding generation in external knowledge, yet they face a fundamental efficiency problem: simple queries consume computational resources equivalent to complex multi-hop reasoning tasks. We present SymRAG, a neuro-symbolic framework that introduces adaptive query routing based on real-time complexity and system load assessments. SymRAG dynamically selects symbolic, neural, or hybrid processing paths to align resource use with query demands. Evaluated on 2,000 queries from HotpotQA and DROP using Llama-3.2-3B and Mistral-7B models, SymRAG achieves 97.6--100.0% exact match accuracy with significantly lower CPU utilization (3.6--6.2%) and processing time (0.985--3.165s). Disabling adaptive logic results in 169--1151% increase in processing time, highlighting the framework's impact. These results underscore the potential of adaptive neuro-symbolic routing for scalable, sustainable AI systems.
Authors: Ethan M. Rudd, Christopher Andrews, Philip Tully
Abstract: Recent advances in generative AI have led to remarkable interest in using systems that rely on large language models (LLMs) for practical applications. However, meaningful evaluation of these systems in real-world scenarios comes with a distinct set of challenges, which are not well-addressed by synthetic benchmarks and de-facto metrics that are often seen in the literature. We present a practical evaluation framework which outlines how to proactively curate representative datasets, select meaningful evaluation metrics, and employ meaningful evaluation methodologies that integrate well with practical development and deployment of LLM-reliant systems that must adhere to real-world requirements and meet user-facing needs.
Authors: Danny Hoang, David Gorsich, Matthew P. Castanier, Farhad Imani
Abstract: Precision process planning in Computer Numerical Control (CNC) machining demands rapid, context-aware decisions on tool selection, feed-speed pairs, and multi-axis routing, placing immense cognitive and procedural burdens on engineers from design specification through final part inspection. Conventional rule-based computer-aided process planning and knowledge-engineering shells freeze domain know-how into static tables, which become limited when dealing with unseen topologies, novel material states, shifting cost-quality-sustainability weightings, or shop-floor constraints such as tool unavailability and energy caps. Large language models (LLMs) promise flexible, instruction-driven reasoning for tasks but they routinely hallucinate numeric values and provide no provenance. We present Augmented Retrieval Knowledge Network Enhanced Search & Synthesis (ARKNESS), the end-to-end framework that fuses zero-shot Knowledge Graph (KG) construction with retrieval-augmented generation to deliver verifiable, numerically exact answers for CNC process planning. ARKNESS (1) automatically distills heterogeneous machining documents, G-code annotations, and vendor datasheets into augmented triple, multi-relational graphs without manual labeling, and (2) couples any on-prem LLM with a retriever that injects the minimal, evidence-linked subgraph needed to answer a query. Benchmarked on 155 industry-curated questions spanning tool sizing and feed-speed optimization, a lightweight 3B-parameter Llama-3 augmented by ARKNESS matches GPT-4o accuracy while achieving a +25 percentage point gain in multiple-choice accuracy, +22.4 pp in F1, and 8.1x ROUGE-L on open-ended responses.
Authors: Joaquin Jordan, Xavier Yin, Melissa Fabros, Gireeja Ranade, Narges Norouzi
Abstract: Automated Essay Scoring (AES) and Automatic Essay Feedback (AEF) systems aim to reduce the workload of human raters in educational assessment. However, most existing systems prioritize numeric scoring accuracy over the quality of feedback. This paper presents Multi-Agent Argumentation and Grammar Integrated Critiquer (MAGIC), a framework that uses multiple specialized agents to evaluate distinct writing aspects to both predict holistic scores and produce detailed, rubric-aligned feedback. To support evaluation, we curated a novel dataset of past GRE practice test essays with expert-evaluated scores and feedback. MAGIC outperforms baseline models in both essay scoring , as measured by Quadratic Weighted Kappa (QWK). We find that despite the improvement in QWK, there are opportunities for future work in aligning LLM-generated feedback to human preferences.
Authors: Haibo Qiu, Xiaohan Lan, Fanfan Liu, Xiaohu Sun, Delian Ruan, Peng Shi, Lin Ma
Abstract: Recent advancements in large language models (LLMs) have witnessed a surge in the development of advanced reasoning paradigms, which are now being integrated into multimodal large language models (MLLMs). However, existing approaches often fall short: methods solely employing reinforcement learning (RL) can struggle with sample inefficiency and activating entirely absent reasoning capabilities, while conventional pipelines that initiate with a cold-start supervised fine-tuning (SFT) phase before RL may restrict the model's exploratory capacity and face suboptimal convergence. In this work, we introduce \textbf{Metis-RISE} (\textbf{R}L \textbf{I}ncentivizes and \textbf{S}FT \textbf{E}nhances) for multimodal reasoning model learning. Unlike conventional approaches, Metis-RISE distinctively omits an initial SFT stage, beginning instead with an RL phase (e.g., using a Group Relative Policy Optimization variant) to incentivize and activate the model's latent reasoning capacity. Subsequently, the targeted SFT stage addresses two key challenges identified during RL: (1) \textit{inefficient trajectory sampling} for tasks where the model possesses but inconsistently applies correct reasoning, which we tackle using self-distilled reasoning trajectories from the RL model itself; and (2) \textit{fundamental capability absence}, which we address by injecting expert-augmented knowledge for prompts where the model entirely fails. This strategic application of RL for incentivization followed by SFT for enhancement forms the core of Metis-RISE, leading to two versions of our MLLMs (7B and 72B parameters). Evaluations on the OpenCompass Multimodal Reasoning Leaderboard demonstrate that both models achieve state-of-the-art performance among similar-sized models, with the 72B version ranking fourth overall.
Authors: Chirag Agarwal
Abstract: While multimodal AI systems (models jointly trained on heterogeneous data types such as text, time series, graphs, and images) have become ubiquitous and achieved remarkable performance across high-stakes applications, transparent and accurate explanation algorithms are crucial for their safe deployment and ensure user trust. However, most existing explainability techniques remain unimodal, generating modality-specific feature attributions, concepts, or circuit traces in isolation and thus failing to capture cross-modal interactions. This paper argues that such unimodal explanations systematically misrepresent and fail to capture the cross-modal influence that drives multimodal model decisions, and the community should stop relying on them for interpreting multimodal models. To support our position, we outline key principles for multimodal explanations grounded in modality: Granger-style modality influence (controlled ablations to quantify how removing one modality changes the explanation for another), Synergistic faithfulness (explanations capture the model's predictive power when modalities are combined), and Unified stability (explanations remain consistent under small, cross-modal perturbations). This targeted shift to multimodal explanations will help the community uncover hidden shortcuts, mitigate modality bias, improve model reliability, and enhance safety in high-stakes settings where incomplete explanations can have serious consequences.
Authors: Daniel Kilov, Caroline Hendy, Secil Yanik Guyot, Aaron J. Snoswell, Seth Lazar
Abstract: Moral competence is the ability to act in accordance with moral principles. As large language models (LLMs) are increasingly deployed in situations demanding moral competence, there is increasing interest in evaluating this ability empirically. We review existing literature and identify three significant shortcoming: (i) Over-reliance on prepackaged moral scenarios with explicitly highlighted moral features; (ii) Focus on verdict prediction rather than moral reasoning; and (iii) Inadequate testing of models' (in)ability to recognize when additional information is needed. Grounded in philosophical research on moral skill, we then introduce a novel method for assessing moral competence in LLMs. Our approach moves beyond simple verdict comparisons to evaluate five dimensions of moral competence: identifying morally relevant features, weighting their importance, assigning moral reasons to these features, synthesizing coherent moral judgments, and recognizing information gaps. We conduct two experiments comparing six leading LLMs against non-expert humans and professional philosophers. In our first experiment using ethical vignettes standard to existing work, LLMs generally outperformed non-expert humans across multiple dimensions of moral reasoning. However, our second experiment, featuring novel scenarios designed to test moral sensitivity by embedding relevant features among irrelevant details, revealed a striking reversal: several LLMs performed significantly worse than humans. Our findings suggest that current evaluations may substantially overestimate LLMs' moral reasoning capabilities by eliminating the task of discerning moral relevance from noisy information, which we take to be a prerequisite for genuine moral skill. This work provides a more nuanced framework for assessing AI moral competence and highlights important directions for improving moral competence in advanced AI systems.
Authors: Qionghao Huang, Lingnuo Lu, Xuemei Wu, Fan Jiang, Xizhe Wang, Xun Wang
Abstract: Adaptive Curriculum Sequencing (ACS) is essential for personalized online learning, yet current approaches struggle to balance complex educational constraints and maintain optimization stability. This paper proposes a Memetic Walrus Optimizer (MWO) that enhances optimization performance through three key innovations: (1) an expert-guided strategy with aging mechanism that improves escape from local optima; (2) an adaptive control signal framework that dynamically balances exploration and exploitation; and (3) a three-tier priority mechanism for generating educationally meaningful sequences. We formulate ACS as a multi-objective optimization problem considering concept coverage, time constraints, and learning style compatibility. Experiments on the OULAD dataset demonstrate MWO's superior performance, achieving 95.3% difficulty progression rate (compared to 87.2% in baseline methods) and significantly better convergence stability (standard deviation of 18.02 versus 28.29-696.97 in competing algorithms). Additional validation on benchmark functions confirms MWO's robust optimization capability across diverse scenarios. The results demonstrate MWO's effectiveness in generating personalized learning sequences while maintaining computational efficiency and solution quality.
Authors: Stella C. Dong, James R. Finlay
Abstract: This paper develops a novel multi-agent reinforcement learning (MARL) framework for reinsurance treaty bidding, addressing long-standing inefficiencies in traditional broker-mediated placement processes. We pose the core research question: Can autonomous, learning-based bidding systems improve risk transfer efficiency and outperform conventional pricing approaches in reinsurance markets? In our model, each reinsurer is represented by an adaptive agent that iteratively refines its bidding strategy within a competitive, partially observable environment. The simulation explicitly incorporates institutional frictions including broker intermediation, incumbent advantages, last-look privileges, and asymmetric access to underwriting information. Empirical analysis demonstrates that MARL agents achieve up to 15% higher underwriting profit, 20% lower tail risk (CVaR), and over 25% improvement in Sharpe ratios relative to actuarial and heuristic baselines. Sensitivity tests confirm robustness across hyperparameter settings, and stress testing reveals strong resilience under simulated catastrophe shocks and capital constraints. These findings suggest that MARL offers a viable path toward more transparent, adaptive, and risk-sensitive reinsurance markets. The proposed framework contributes to emerging literature at the intersection of algorithmic market design, strategic bidding, and AI-enabled financial decision-making.
Authors: Alexander Novikov, Ng\^an V\~u, Marvin Eisenberger, Emilien Dupont, Po-Sen Huang, Adam Zsolt Wagner, Sergey Shirobokov, Borislav Kozlovskii, Francisco J. R. Ruiz, Abbas Mehrabian, M. Pawan Kumar, Abigail See, Swarat Chaudhuri, George Holland, Alex Davies, Sebastian Nowozin, Pushmeet Kohli, Matej Balog
Abstract: In this white paper, we present AlphaEvolve, an evolutionary coding agent that substantially enhances capabilities of state-of-the-art LLMs on highly challenging tasks such as tackling open scientific problems or optimizing critical pieces of computational infrastructure. AlphaEvolve orchestrates an autonomous pipeline of LLMs, whose task is to improve an algorithm by making direct changes to the code. Using an evolutionary approach, continuously receiving feedback from one or more evaluators, AlphaEvolve iteratively improves the algorithm, potentially leading to new scientific and practical discoveries. We demonstrate the broad applicability of this approach by applying it to a number of important computational problems. When applied to optimizing critical components of large-scale computational stacks at Google, AlphaEvolve developed a more efficient scheduling algorithm for data centers, found a functionally equivalent simplification in the circuit design of hardware accelerators, and accelerated the training of the LLM underpinning AlphaEvolve itself. Furthermore, AlphaEvolve discovered novel, provably correct algorithms that surpass state-of-the-art solutions on a spectrum of problems in mathematics and computer science, significantly expanding the scope of prior automated discovery methods (Romera-Paredes et al., 2023). Notably, AlphaEvolve developed a search algorithm that found a procedure to multiply two $4 \times 4$ complex-valued matrices using $48$ scalar multiplications; offering the first improvement, after 56 years, over Strassen's algorithm in this setting. We believe AlphaEvolve and coding agents like it can have a significant impact in improving solutions of problems across many areas of science and computation.
Authors: Theofanis Aravanis
Abstract: Artificial Neural Networks (ANNs) are powerful machine-learning models capable of capturing intricate non-linear relationships. They are widely used nowadays across numerous scientific and engineering domains, driving advancements in both research and real-world applications. In our recent work, we focused on the statics and dynamics of a particular subclass of ANNs, which we refer to as binary ANNs. A binary ANN is a feed-forward network in which both inputs and outputs are restricted to binary values, making it particularly suitable for a variety of practical use cases. Our previous study approached binary ANNs through the lens of belief-change theory, specifically the Alchourron, Gardenfors and Makinson (AGM) framework, yielding several key insights. Most notably, we demonstrated that the knowledge embodied in a binary ANN (expressed through its input-output behaviour) can be symbolically represented using a propositional logic language. Moreover, the process of modifying a belief set (through revision or contraction) was mapped onto a gradual transition through a series of intermediate belief sets. Analogously, the training of binary ANNs was conceptualized as a sequence of such belief-set transitions, which we showed can be formalized using full-meet AGM-style belief change. In the present article, we extend this line of investigation by addressing some critical limitations of our previous study. Specifically, we show that Dalal's method for belief change naturally induces a structured, gradual evolution of states of belief. More importantly, given the known shortcomings of full-meet belief change, we demonstrate that the training dynamics of binary ANNs can be more effectively modelled using robust AGM-style change operations -- namely, lexicographic revision and moderate contraction -- that align with the Darwiche-Pearl framework for iterated belief change.
Authors: Steve Yuwono, Muhammad Uzair Rana, Dorothea Schwung, Andreas Schwung
Abstract: This paper presents a novel method for enhancing the adaptability of Proportional-Integral-Derivative (PID) controllers in industrial systems using event-based dynamic game theory, which enables the PID controllers to self-learn, optimize, and fine-tune themselves. In contrast to conventional self-learning approaches, our proposed framework offers an event-driven control strategy and game-theoretic learning algorithms. The players collaborate with the PID controllers to dynamically adjust their gains in response to set point changes and disturbances. We provide a theoretical analysis showing sound convergence guarantees for the game given suitable stability ranges of the PID controlled loop. We further introduce an automatic boundary detection mechanism, which helps the players to find an optimal initialization of action spaces and significantly reduces the exploration time. The efficacy of this novel methodology is validated through its implementation in the temperature control loop of a printing press machine. Eventually, the outcomes of the proposed intelligent self-tuning PID controllers are highly promising, particularly in terms of reducing overshoot and settling time.
Authors: Zhenyu Xia, Xinlei Huang, Suvash C. Saha
Abstract: Electroencephalography (EEG) is extensively employed in medical diagnostics and brain-computer interface (BCI) applications due to its non-invasive nature and high temporal resolution. However, EEG analysis faces significant challenges, including noise, nonstationarity, and inter-subject variability, which hinder its clinical utility. Traditional neural networks often lack integration with biophysical knowledge, limiting their interpretability, robustness, and potential for medical translation. To address these limitations, this study introduces NeuroPhysNet, a novel Physics-Informed Neural Network (PINN) framework tailored for EEG signal analysis and motor imagery classification in medical contexts. NeuroPhysNet incorporates the FitzHugh-Nagumo model, embedding neurodynamical principles to constrain predictions and enhance model robustness. Evaluated on the BCIC-IV-2a dataset, the framework achieved superior accuracy and generalization compared to conventional methods, especially in data-limited and cross-subject scenarios, which are common in clinical settings. By effectively integrating biophysical insights with data-driven techniques, NeuroPhysNet not only advances BCI applications but also holds significant promise for enhancing the precision and reliability of clinical diagnostics, such as motor disorder assessments and neurorehabilitation planning.
Authors: Jakub Kowalski, Mark H. M. Winands, Maksymilian Wi\'sniewski, Stanis{\l}aw Reda, Anna Wilbik
Abstract: Typically, research on Explainable Artificial Intelligence (XAI) focuses on black-box models within the context of a general policy in a known, specific domain. This paper advocates for the need for knowledge-agnostic explainability applied to the subfield of XAI called Explainable Search, which focuses on explaining the choices made by intelligent search techniques. It proposes Monte-Carlo Tree Search (MCTS) enhancements as a solution to obtaining additional data and providing higher-quality explanations while remaining knowledge-free, and analyzes the most popular enhancements in terms of the specific types of explainability they introduce. So far, no other research has considered the explainability of MCTS enhancements. We present a proof-of-concept that demonstrates the advantages of utilizing enhancements.
Authors: Guoxi Zhang, Jiawei Chen, Tianzhuo Yang, Jiaming Ji, Yaodong Yang, Juntao Dai
Abstract: The increasing prevalence of large language models (LLMs) is influencing global value systems. However, these models frequently exhibit a pronounced WEIRD (Western, Educated, Industrialized, Rich, Democratic) cultural bias due to lack of attention to minority values. This monocultural perspective may reinforce dominant values and marginalize diverse cultural viewpoints, posing challenges for the development of equitable and inclusive AI systems. In this work, we introduce a systematic framework designed to boost fair and robust cross-cultural consensus among LLMs. We model consensus as a Nash Equilibrium and employ a game-theoretic negotiation method based on Policy-Space Response Oracles (PSRO) to simulate an organized cross-cultural negotiation process. To evaluate this approach, we construct regional cultural agents using data transformed from the World Values Survey (WVS). Beyond the conventional model-level evaluation method, We further propose two quantitative metrics, Perplexity-based Acceptence and Values Self-Consistency, to assess consensus outcomes. Experimental results indicate that our approach generates consensus of higher quality while ensuring more balanced compromise compared to baselines. Overall, it mitigates WEIRD bias by guiding agents toward convergence through fair and gradual negotiation steps.
Authors: Jakub Kowalski, Dennis J. N. J. Soemers, Szymon Kosakowski, Mark H. M. Winands
Abstract: This paper presents Generalized Proof-Number Monte-Carlo Tree Search: a generalization of recently proposed combinations of Proof-Number Search (PNS) with Monte-Carlo Tree Search (MCTS), which use (dis)proof numbers to bias UCB1-based Selection strategies towards parts of the search that are expected to be easily (dis)proven. We propose three core modifications of prior combinations of PNS with MCTS. First, we track proof numbers per player. This reduces code complexity in the sense that we no longer need disproof numbers, and generalizes the technique to be applicable to games with more than two players. Second, we propose and extensively evaluate different methods of using proof numbers to bias the selection strategy, achieving strong performance with strategies that are simpler to implement and compute. Third, we merge our technique with Score Bounded MCTS, enabling the algorithm to prove and leverage upper and lower bounds on scores - as opposed to only proving wins or not-wins. Experiments demonstrate substantial performance increases, reaching the range of 80% for 8 out of the 11 tested board games.
Authors: Kaspar Rothenfusser, Bekk Blando
Abstract: Large Language Models (LLMs) possess intricate internal representations of the world, yet these latent structures are notoriously difficult to interpret or repurpose beyond the original prediction task. Building on our earlier work (Rothenfusser, 2025), which introduced the concept of vector ontologies as a framework for translating high-dimensional neural representations into interpretable geometric structures, this paper provides the first empirical validation of that approach. A vector ontology defines a domain-specific vector space spanned by ontologically meaningful dimensions, allowing geometric analysis of concepts and relationships within a domain. We construct an 8-dimensional vector ontology of musical genres based on Spotify audio features and test whether an LLM's internal world model of music can be consistently and accurately projected into this space. Using GPT-4o-mini, we extract genre representations through multiple natural language prompts and analyze the consistency of these projections across linguistic variations and their alignment with ground-truth data. Our results show (1) high spatial consistency of genre projections across 47 query formulations, (2) strong alignment between LLM-inferred genre locations and real-world audio feature distributions, and (3) evidence of a direct relationship between prompt phrasing and spatial shifts in the LLM's inferred vector ontology. These findings demonstrate that LLMs internalize structured, repurposable knowledge and that vector ontologies offer a promising method for extracting and analyzing this knowledge in a transparent and verifiable way.
Authors: Yuefei Lyu, Chaozhuo Li, Xi Zhang, Tianle Zhang
Abstract: Text-attributed graphs (TAGs) integrate textual data with graph structures, providing valuable insights in applications such as social network analysis and recommendation systems. Graph Neural Networks (GNNs) effectively capture both topological structure and textual information in TAGs but are vulnerable to adversarial attacks. Existing graph injection attack (GIA) methods assume that attackers can directly manipulate the embedding layer, producing non-explainable node embeddings. Furthermore, the effectiveness of these attacks often relies on surrogate models with high training costs. Thus, this paper introduces ATAG-LLM, a novel black-box GIA framework tailored for TAGs. Our approach leverages large language models (LLMs) to generate interpretable text-level node attributes directly, ensuring attacks remain feasible in real-world scenarios. We design strategies for LLM prompting that balance exploration and reliability to guide text generation, and propose a similarity assessment method to evaluate attack text effectiveness in disrupting graph homophily. This method efficiently perturbs the target node with minimal training costs in a strict black-box setting, ensuring a text-level graph injection attack for TAGs. Experiments on real-world TAG datasets validate the superior performance of ATAG-LLM compared to state-of-the-art embedding-level and text-level attack methods.
Authors: Gianni Molinari, Fabio Ciravegna
Abstract: The rapid advancement of intelligent agents and Large Language Models (LLMs) is reshaping the pervasive computing field. Their ability to perceive, reason, and act through natural language understanding enables autonomous problem-solving in complex pervasive environments, including the management of heterogeneous sensors, devices, and data. This survey outlines the architectural components of LLM agents (profiling, memory, planning, and action) and examines their deployment and evaluation across various scenarios. Than it reviews computational and infrastructural advancements (cloud to edge) in pervasive computing and how AI is moving in this field. It highlights state-of-the-art agent deployment strategies and applications, including local and distributed execution on resource-constrained devices. This survey identifies key challenges of these agents in pervasive computing such as architectural, energetic and privacy limitations. It finally proposes what we called "Agent as a Tool", a conceptual framework for pervasive agentic AI, emphasizing context awareness, modularity, security, efficiency and effectiveness.
Authors: Zhen Yao, Elisabetta De Maria, Robert De Simone
Abstract: Spiking Neural Networks (SNN) are models for "realistic" neuronal computation, which makes them somehow different in scope from "ordinary" deep-learning models widely used in AI platforms nowadays. SNNs focus on timed latency (and possibly probability) of neuronal reactive activation/response, more than numerical computation of filters. So, an SNN model must provide modeling constructs for elementary neural bundles and then for synaptic connections to assemble them into compound data flow network patterns. These elements are to be parametric patterns, with latency and probability values instantiated on particular instances (while supposedly constant "at runtime"). Designers could also use different values to represent "tired" neurons, or ones impaired by external drugs, for instance. One important challenge in such modeling is to study how compound models could meet global reaction requirements (in stochastic timing challenges), provided similar provisions on individual neural bundles. A temporal language of logic to express such assume/guarantee contracts is thus needed. This may lead to formal verification on medium-sized models and testing observations on large ones. In the current article, we make preliminary progress at providing a simple model framework to express both elementary SNN neural bundles and their connecting constructs, which translates readily into both a model-checker and a simulator (both already existing and robust) to conduct experiments.
Authors: Wooseok Seo, Seungju Han, Jaehun Jung, Benjamin Newman, Seungwon Lim, Seungbeen Lee, Ximing Lu, Yejin Choi, Youngjae Yu
Abstract: Fact verification is essential for ensuring the reliability of LLM applications. In this study, we evaluate 12 pre-trained LLMs and one specialized fact-verifier, including frontier LLMs and open-weight reasoning LLMs, using a collection of examples from 14 fact-checking benchmarks. We share three findings intended to guide future development of more robust fact verifiers. First, we highlight the importance of addressing annotation errors and ambiguity in datasets, demonstrating that approximately 16\% of ambiguous or incorrectly labeled data substantially influences model rankings. Neglecting this issue may result in misleading conclusions during comparative evaluations, and we suggest using a systematic pipeline utilizing LLM-as-a-judge to help identify these issues at scale. Second, we discover that frontier LLMs with few-shot in-context examples, often overlooked in previous works, achieve top-tier performance. We therefore recommend future studies include comparisons with these simple yet highly effective baselines. Lastly, despite their effectiveness, frontier LLMs incur substantial costs, motivating the development of small, fine-tuned fact verifiers. We show that these small models still have room for improvement, particularly on instances that require complex reasoning. Encouragingly, we demonstrate that augmenting training with synthetic multi-hop reasoning data significantly enhances their capabilities in such instances. We release our code, model, and dataset at https://github.com/just1nseo/verifying-the-verifiers
Authors: Xiangfan Wu
Abstract: Current Reinforcement Learning (RL) methodologies for Large Language Models (LLMs) often rely on simplistic, outcome-based reward signals (e.g., final answer correctness), which limits the depth of learning from each interaction. This paper introduces Socratic Reinforcement Learning (Socratic-RL), a novel, process-oriented framework designed to address this limitation. Socratic-RL operates on the principle that deeper understanding is achieved by reflecting on the causal reasons for errors and successes within the reasoning process itself. The framework employs a decoupled "Teacher-Student" architecture, where a "Teacher AI" analyzes interaction histories, extracts causal insights, and formulates them into structured "viewpoints." These viewpoints, acting as distilled guidance, are then used by a "Student AI" to enhance its subsequent reasoning. A key innovation is the iterative self-improvement of the Teacher AI, enabling its reflective capabilities to evolve through a meta-learning loop. To manage the accumulation of knowledge, a distillation mechanism compresses learned viewpoints into the Student's parameters. By focusing on process rather than just outcome, Socratic-RL presents a pathway toward enhanced sample efficiency, superior interpretability, and a more scalable architecture for self-improving AI systems. This paper details the foundational concepts, formal mechanisms, synergies, challenges, and a concrete research roadmap for this proposed framework.
Authors: Leonie V. D. E. Vogelsmeier, Eduardo Oliveira, Kamila Misiejuk, Sonsoles L\'opez-Pernas, Mohammed Saqr
Abstract: Large language models (LLMs) offer the potential to simulate human-like responses and behaviors, creating new opportunities for psychological science. In the context of self-regulated learning (SRL), if LLMs can reliably simulate survey responses at scale and speed, they could be used to test intervention scenarios, refine theoretical models, augment sparse datasets, and represent hard-to-reach populations. However, the validity of LLM-generated survey responses remains uncertain, with limited research focused on SRL and existing studies beyond SRL yielding mixed results. Therefore, in this study, we examined LLM-generated responses to the 44-item Motivated Strategies for Learning Questionnaire (MSLQ; Pintrich \& De Groot, 1990), a widely used instrument assessing students' learning strategies and academic motivation. Particularly, we used the LLMs GPT-4o, Claude 3.7 Sonnet, Gemini 2 Flash, LLaMA 3.1-8B, and Mistral Large. We analyzed item distributions, the psychological network of the theoretical SRL dimensions, and psychometric validity based on the latent factor structure. Our results suggest that Gemini 2 Flash was the most promising LLM, showing considerable sampling variability and producing underlying dimensions and theoretical relationships that align with prior theory and empirical findings. At the same time, we observed discrepancies and limitations, underscoring both the potential and current constraints of using LLMs for simulating psychological survey data and applying it in educational contexts.
Authors: Alex Grzankowski, Geoff Keeling, Henry Shevlin, Winnie Street
Abstract: Many people feel compelled to interpret, describe, and respond to Large Language Models (LLMs) as if they possess inner mental lives similar to our own. Responses to this phenomenon have varied. Inflationists hold that at least some folk psychological ascriptions to LLMs are warranted. Deflationists argue that all such attributions of mentality to LLMs are misplaced, often cautioning against the risk that anthropomorphic projection may lead to misplaced trust or potentially even confusion about the moral status of LLMs. We advance this debate by assessing two common deflationary arguments against LLM mentality. What we term the 'robustness strategy' aims to undercut one justification for believing that LLMs are minded entities by showing that putatively cognitive and humanlike behaviours are not robust, failing to generalise appropriately. What we term the 'etiological strategy' undercuts attributions of mentality by challenging naive causal explanations of LLM behaviours, offering alternative causal accounts that weaken the case for mental state attributions. While both strategies offer powerful challenges to full-blown inflationism, we find that neither strategy provides a knock-down case against ascriptions of mentality to LLMs simpliciter. With this in mind, we explore a modest form of inflationism that permits ascriptions of mentality to LLMs under certain conditions. Specifically, we argue that folk practice provides a defeasible basis for attributing mental states and capacities to LLMs provided those mental states and capacities can be understood in metaphysically undemanding terms (e.g. knowledge, beliefs and desires), while greater caution is required when attributing metaphysically demanding mental phenomena such as phenomenal consciousness.
Authors: Xialie Zhuang, Peixian Ma, Zhikai Jia, Zheng Cao, Shiwei Liu
Abstract: The ongoing evolution of language models has led to the development of large-scale architectures that demonstrate exceptional performance across a wide range of tasks. However, these models come with significant computational and energy demands, as well as potential privacy implications. In this context, Small Reasoning Language Models (SRLMs) with approximately 0.5 billion parameters present a compelling alternative due to their remarkable computational efficiency and cost effectiveness, particularly in resource-constrained environments. Despite these advantages, the limited capacity of 0.5 billion parameter models poses challenges in handling complex tasks such as mathematical reasoning and code generation. This research investigates various training strategies, including supervised fine-tuning (SFT), knowledge distillation (KD), and reinforcement learning (RL), as well as their hybrid implementations, to enhance the performance of 0.5B SRLMs. We analyze effective methodologies to bridge the performance gap between SRLMS and larger models and present insights into optimal training pipelines tailored for these smaller architectures. Through extensive experimental validation and analysis, our work aims to provide actionable recommendations for maximizing the reasoning capabilities of 0.5B models.
Authors: Kangye Ji, Yuan Meng, Hanyun Cui, Ye Li, Shengjia Hua, Lei Chen, Zhi Wang
Abstract: Diffusion Policy has demonstrated strong visuomotor modeling capabilities, but its high computational cost renders it impractical for real-time robotic control. Despite huge redundancy across repetitive denoising steps, existing diffusion acceleration techniques fail to generalize to Diffusion Policy due to fundamental architectural and data divergences. In this paper, we propose Block-wise Adaptive Caching(BAC), a method to accelerate Diffusion Policy by caching intermediate action features. BAC achieves lossless action generation acceleration by adaptively updating and reusing cached features at the block level, based on a key observation that feature similarities vary non-uniformly across timesteps and locks. To operationalize this insight, we first propose the Adaptive Caching Scheduler, designed to identify optimal update timesteps by maximizing the global feature similarities between cached and skipped features. However, applying this scheduler for each block leads to signiffcant error surges due to the inter-block propagation of caching errors, particularly within Feed-Forward Network (FFN) blocks. To mitigate this issue, we develop the Bubbling Union Algorithm, which truncates these errors by updating the upstream blocks with signiffcant caching errors before downstream FFNs. As a training-free plugin, BAC is readily integrable with existing transformer-based Diffusion Policy and vision-language-action models. Extensive experiments on multiple robotic benchmarks demonstrate that BAC achieves up to 3x inference speedup for free.
Authors: Daniel Anadria, Roel Dobbe, Anastasia Giachanou, Ruurd Kuiper, Richard Bartels, \'I\~nigo Mart\'inez de Rituerto de Troya, Carmen Z\"urcher, Daniel Oberski
Abstract: In this work, we reflect on the data-driven modeling paradigm that is gaining ground in AI-driven automation of patient care. We argue that the repurposing of existing real-world patient datasets for machine learning may not always represent an optimal approach to model development as it could lead to undesirable outcomes in patient care. We reflect on the history of data analysis to explain how the data-driven paradigm rose to popularity, and we envision ways in which systems thinking and clinical domain theory could complement the existing model development approaches in reaching human-centric outcomes. We call for a purpose-driven machine learning paradigm that is grounded in clinical theory and the sociotechnical realities of real-world operational contexts. We argue that understanding the utility of existing patient datasets requires looking in two directions: upstream towards the data generation, and downstream towards the automation objectives. This purpose-driven perspective to AI system development opens up new methodological opportunities and holds promise for AI automation of patient care.
Authors: Ken Huang, Akram Sheriff, Vineeth Sai Narajala, Idan Habler
Abstract: As multi-agent systems evolve to encompass increasingly diverse and specialized agents, the challenge of enabling effective collaboration between heterogeneous agents has become paramount, with traditional agent communication protocols often assuming homogeneous environments or predefined interaction patterns that limit their applicability in dynamic, open-world scenarios. This paper presents the Agent Capability Negotiation and Binding Protocol (ACNBP), a novel framework designed to facilitate secure, efficient, and verifiable interactions between agents in heterogeneous multi-agent systems through integration with an Agent Name Service (ANS) infrastructure that provides comprehensive discovery, negotiation, and binding mechanisms. The protocol introduces a structured 10-step process encompassing capability discovery, candidate pre-screening and selection, secure negotiation phases, and binding commitment with built-in security measures including digital signatures, capability attestation, and comprehensive threat mitigation strategies, while a key innovation of ACNBP is its protocolExtension mechanism that enables backward-compatible protocol evolution and supports diverse agent architectures while maintaining security and interoperability. We demonstrate ACNBP's effectiveness through a comprehensive security analysis using the MAESTRO threat modeling framework, practical implementation considerations, and a detailed example showcasing the protocol's application in a document translation scenario, with the protocol addressing critical challenges in agent autonomy, capability verification, secure communication, and scalable agent ecosystem management.
Authors: Hidetomo Nabeshima, Mutsunori Banbara, Torsten Schaub, Takehide Soh
Abstract: We present the design principles of a nurse scheduling system built using Answer Set Programming (ASP) and successfully deployed at the University of Yamanashi Hospital. Nurse scheduling is a complex optimization problem requiring the reconciliation of individual nurse preferences with hospital staffing needs across various wards. This involves balancing hard and soft constraints and the flexibility of interactive adjustments. While extensively studied in academia, real-world nurse scheduling presents unique challenges that go beyond typical benchmark problems and competitions. This paper details the practical application of ASP to address these challenges at the University of Yamanashi Hospital, focusing on the insights gained and the advancements in ASP technology necessary to effectively manage the complexities of real-world deployment.
Authors: Jonah Brown-Cohen, Geoffrey Irving, Georgios Piliouras
Abstract: Training powerful AI systems to exhibit desired behaviors hinges on the ability to provide accurate human supervision on increasingly complex tasks. A promising approach to this problem is to amplify human judgement by leveraging the power of two competing AIs in a debate about the correct solution to a given problem. Prior theoretical work has provided a complexity-theoretic formalization of AI debate, and posed the problem of designing protocols for AI debate that guarantee the correctness of human judgements for as complex a class of problems as possible. Recursive debates, in which debaters decompose a complex problem into simpler subproblems, hold promise for growing the class of problems that can be accurately judged in a debate. However, existing protocols for recursive debate run into the obfuscated arguments problem: a dishonest debater can use a computationally efficient strategy that forces an honest opponent to solve a computationally intractable problem to win. We mitigate this problem with a new recursive debate protocol that, under certain stability assumptions, ensures that an honest debater can win with a strategy requiring computational efficiency comparable to their opponent.
Authors: Shaolei Zhang, Shoutao Guo, Qingkai Fang, Yan Zhou, Yang Feng
Abstract: The emergence of GPT-4o-like large multimodal models (LMMs) has raised the exploration of integrating text, vision, and speech modalities to support more flexible multimodal interaction. Existing LMMs typically concatenate representation of modalities along the sequence dimension and feed them into a large language model (LLM) backbone. While sequence-dimension concatenation is straightforward for modality integration, it often relies heavily on large-scale data to learn modality alignments. In this paper, we aim to model the relationships between modalities more purposefully, thereby achieving more efficient and flexible modality alignments. To this end, we propose Stream-Omni, a large language-vision-speech model with efficient modality alignments, which can simultaneously support interactions under various modality combinations. Stream-Omni employs LLM as the backbone and aligns the vision and speech to the text based on their relationships. For vision that is semantically complementary to text, Stream-Omni uses sequence-dimension concatenation to achieve vision-text alignment. For speech that is semantically consistent with text, Stream-Omni introduces a CTC-based layer-dimension mapping to achieve speech-text alignment. In this way, Stream-Omni can achieve modality alignments with less data (especially speech), enabling the transfer of text capabilities to other modalities. Experiments on various benchmarks demonstrate that Stream-Omni achieves strong performance on visual understanding, speech interaction, and vision-grounded speech interaction tasks. Owing to the layer-dimensional mapping, Stream-Omni can simultaneously provide intermediate text outputs (such as ASR transcriptions and model responses) during speech interaction, offering users a comprehensive multimodal experience.
Authors: Arjun Krishna, Aaditya Rastogi, Erick Galinkin
Abstract: The introduction of advanced reasoning capabilities have improved the problem-solving performance of large language models, particularly on math and coding benchmarks. However, it remains unclear whether these reasoning models are more or less vulnerable to adversarial prompt attacks than their non-reasoning counterparts. In this work, we present a systematic evaluation of weaknesses in advanced reasoning models compared to similar non-reasoning models across a diverse set of prompt-based attack categories. Using experimental data, we find that on average the reasoning-augmented models are \emph{slightly more robust} than non-reasoning models (42.51\% vs 45.53\% attack success rate, lower is better). However, this overall trend masks significant category-specific differences: for certain attack types the reasoning models are substantially \emph{more vulnerable} (e.g., up to 32 percentage points worse on a tree-of-attacks prompt), while for others they are markedly \emph{more robust} (e.g., 29.8 points better on cross-site scripting injection). Our findings highlight the nuanced security implications of advanced reasoning in language models and emphasize the importance of stress-testing safety across diverse adversarial techniques.
Authors: Brahim Driss, Alex Davey, Riad Akrour
Abstract: Preference-based reinforcement learning (PbRL) has emerged as a promising approach for learning behaviors from human feedback without predefined reward functions. However, current PbRL methods face a critical challenge in effectively exploring the preference space, often converging prematurely to suboptimal policies that satisfy only a narrow subset of human preferences. In this work, we identify and address this preference exploration problem through population-based methods. We demonstrate that maintaining a diverse population of agents enables more comprehensive exploration of the preference landscape compared to single-agent approaches. Crucially, this diversity improves reward model learning by generating preference queries with clearly distinguishable behaviors, a key factor in real-world scenarios where humans must easily differentiate between options to provide meaningful feedback. Our experiments reveal that current methods may fail by getting stuck in local optima, requiring excessive feedback, or degrading significantly when human evaluators make errors on similar trajectories, a realistic scenario often overlooked by methods relying on perfect oracle teachers. Our population-based approach demonstrates robust performance when teachers mislabel similar trajectory segments and shows significantly enhanced preference exploration capabilities,particularly in environments with complex reward landscapes.
Authors: Murat Kirisci, Nihat Topac, Musa Bardak
Abstract: There are many genetic and environmental factors that affect cognitive development. Music education can also be considered as one of the environmental factors. Some researchers emphasize that Music is an action that requires meta-cognitive functions such as mathematics and chess and supports spatial intelligence. The effect of Music on cognitive development in early childhood was examined using the Pythagorean Fuzzy Sets(PFS) method defined by Yager. This study created PFS based on experts' opinions, and an algorithm was given according to PFS. The algorithm's results supported the experts' data on the development of spatial-temporal skills in music education given in early childhood. The algorithm's ranking was done using the Expectation Score Function. The rankings obtained from the algorithm overlap with the experts' rankings.
Authors: Oliver Broadrick, Sanyam Agarwal, Guy Van den Broeck, Markus Bl\"aser
Abstract: Marginalization -- summing a function over all assignments to a subset of its inputs -- is a fundamental computational problem with applications from probabilistic inference to formal verification. Despite its computational hardness in general, there exist many classes of functions (e.g., probabilistic models) for which marginalization remains tractable, and they can be commonly expressed by polynomial size arithmetic circuits computing multilinear polynomials. This raises the question, can all functions with polynomial time marginalization algorithms be succinctly expressed by such circuits? We give a negative answer, exhibiting simple functions with tractable marginalization yet no efficient representation by known models, assuming $\textsf{FP}\neq\#\textsf{P}$ (an assumption implied by $\textsf{P} \neq \textsf{NP}$). To this end, we identify a hierarchy of complexity classes corresponding to stronger forms of marginalization, all of which are efficiently computable on the known circuit models. We conclude with a completeness result, showing that whenever there is an efficient real RAM performing virtual evidence marginalization for a function, then there are small circuits for that function's multilinear representation.
Authors: Md Mahbub Alam, Amilcar Soares, Jos\'e F. Rodrigues-Jr, Gabriel Spadon
Abstract: Accurate vessel trajectory prediction is crucial for navigational safety, route optimization, traffic management, search and rescue operations, and autonomous navigation. Traditional data-driven models lack real-world physical constraints, leading to forecasts that disobey vessel motion dynamics, such as in scenarios with limited or noisy data where sudden course changes or speed variations occur due to external factors. To address this limitation, we propose a Physics-Informed Neural Network (PINN) approach for trajectory prediction that integrates a streamlined kinematic model for vessel motion into the neural network training process via a first- and second-order, finite difference physics-based loss function. This loss function, discretized using the first-order forward Euler method, Heun's second-order approximation, and refined with a midpoint approximation based on Taylor series expansion, enforces fidelity to fundamental physical principles by penalizing deviations from expected kinematic behavior. We evaluated PINN using real-world AIS datasets that cover diverse maritime conditions and compared it with state-of-the-art models. Our results demonstrate that the proposed method reduces average displacement errors by up to 32% across models and datasets while maintaining physical consistency. These results enhance model reliability and adherence to mission-critical maritime activities, where precision translates into better situational awareness in the oceans.
Authors: Md. Biplob Hosen, Sabbir Ahmed, Bushra Akter, Mehrin Anannya
Abstract: Understanding socio-academic and economic factors influencing students' performance is crucial for effective educational interventions. This study employs several machine learning techniques and causal analysis to predict and elucidate the impacts of these factors on academic performance. We constructed a hypothetical causal graph and collected data from 1,050 student profiles. Following meticulous data cleaning and visualization, we analyze linear relationships through correlation and variable plots, and perform causal analysis on the hypothetical graph. Regression and classification models are applied for prediction, and unsupervised causality analysis using PC, GES, ICA-LiNGAM, and GRASP algorithms is conducted. Our regression analysis shows that Ridge Regression achieve a Mean Absolute Error (MAE) of 0.12 and a Mean Squared Error (MSE) of 0.024, indicating robustness, while classification models like Random Forest achieve nearly perfect F1-scores. The causal analysis shows significant direct and indirect effects of factors such as class attendance, study hours, and group study on CGPA. These insights are validated through unsupervised causality analysis. By integrating the best regression model into a web application, we are developing a practical tool for students and educators to enhance academic outcomes based on empirical evidence.
Authors: Minh-Duong Nguyen, Le-Tuan Nguyen, Quoc-Viet Pham
Abstract: Federated Continual Learning (FCL) has recently emerged as a crucial research area, as data from distributed clients typically arrives as a stream, requiring sequential learning. This paper explores a more practical and challenging FCL setting, where clients may have unrelated or even conflicting data and tasks. In this scenario, statistical heterogeneity and data noise can create spurious correlations, leading to biased feature learning and catastrophic forgetting. Existing FCL approaches often use generative replay to create pseudo-datasets of previous tasks. However, generative replay itself suffers from catastrophic forgetting and task divergence among clients, leading to overfitting in FCL. Existing FCL approaches often use generative replay to create pseudo-datasets of previous tasks. However, generative replay itself suffers from catastrophic forgetting and task divergence among clients, leading to overfitting in FCL. To address these challenges, we propose a novel approach called Spatio-Temporal grAdient Matching with network-free Prototype (STAMP). Our contributions are threefold: 1) We develop a model-agnostic method to determine subset of samples that effectively form prototypes when using a prototypical network, making it resilient to continual learning challenges; 2) We introduce a spatio-temporal gradient matching approach, applied at both the client-side (temporal) and server-side (spatial), to mitigate catastrophic forgetting and data heterogeneity; 3) We leverage prototypes to approximate task-wise gradients, improving gradient matching on the client-side. Extensive experiments demonstrate our method's superiority over existing baselines.
Authors: Krti Tallam
Abstract: We present a robust neural watermarking framework for scientific data integrity, targeting high-dimensional fields common in climate modeling and fluid simulations. Using a convolutional autoencoder, binary messages are invisibly embedded into structured data such as temperature, vorticity, and geopotential. Our method ensures watermark persistence under lossy transformations - including noise injection, cropping, and compression - while maintaining near-original fidelity (sub-1\% MSE). Compared to classical singular value decomposition (SVD)-based watermarking, our approach achieves $>$98\% bit accuracy and visually indistinguishable reconstructions across ERA5 and Navier-Stokes datasets. This system offers a scalable, model-compatible tool for data provenance, auditability, and traceability in high-performance scientific workflows, and contributes to the broader goal of securing AI systems through verifiable, physics-aware watermarking. We evaluate on physically grounded scientific datasets as a representative stress-test; the framework extends naturally to other structured domains such as satellite imagery and autonomous-vehicle perception streams.
Authors: Mayesha Tasnim, Erman Acar, Sennay Ghebreab
Abstract: The design of fair and efficient algorithms for allocating public resources, such as school admissions, housing, or medical residency, has a profound social impact. In one-sided matching problems, where individuals are assigned to items based on ranked preferences, a fundamental trade-off exists between efficiency and strategyproofness. Existing algorithms like Random Serial Dictatorship (RSD), Probabilistic Serial (PS), and Rank Minimization (RM) capture only one side of this trade-off: RSD is strategyproof but inefficient, while PS and RM are efficient but incentivize manipulation. We propose EMERGENT, a novel application of Generative Flow Networks (GFlowNets) to one-sided matching, leveraging its ability to sample diverse, high-reward solutions. In our approach, efficient and manipulation-resistant matches emerge naturally: high-reward solutions yield efficient matches, while the stochasticity of GFlowNets-based outputs reduces incentives for manipulation. Experiments show that EMERGENT outperforms RSD in rank efficiency while significantly reducing strategic vulnerability compared to matches produced by RM and PS. Our work highlights the potential of GFlowNets for applications involving social choice mechanisms, where it is crucial to balance efficiency and manipulability.
Authors: Dylan Kline
Abstract: This study bridges cognitive science and neural network design by examining whether artificial models exhibit human-like forgetting curves. Drawing upon Ebbinghaus' seminal work on memory decay and principles of spaced repetition, we propose a quantitative framework to measure information retention in neural networks. Our approach computes the recall probability by evaluating the similarity between a network's current hidden state and previously stored prototype representations. This retention metric facilitates the scheduling of review sessions, thereby mitigating catastrophic forgetting during deployment and enhancing training efficiency by prompting targeted reviews. Our experiments with Multi-Layer Perceptrons reveal human-like forgetting curves, with knowledge becoming increasingly robust through scheduled reviews. This alignment between neural network forgetting curves and established human memory models identifies neural networks as an architecture that naturally emulates human memory decay and can inform state-of-the-art continual learning algorithms.
Authors: Chaoyi Jiang, Sungwoo Kim, Lei Gao, Hossein Entezari Zarch, Won Woo Ro, Murali Annavaram
Abstract: Masked autoregressive (MAR) models unify the strengths of masked and autoregressive generation by predicting tokens in a fixed order using bidirectional attention for image generation. While effective, MAR models suffer from significant computational overhead, as they recompute attention and feed-forward representations for all tokens at every decoding step, despite most tokens remaining semantically stable across steps. We propose a training-free generation framework MARch\'e to address this inefficiency through two key components: cache-aware attention and selective KV refresh. Cache-aware attention partitions tokens into active and cached sets, enabling separate computation paths that allow efficient reuse of previously computed key/value projections without compromising full-context modeling. But a cached token cannot be used indefinitely without recomputation due to the changing contextual information over multiple steps. MARch\'e recognizes this challenge and applies a technique called selective KV refresh. Selective KV refresh identifies contextually relevant tokens based on attention scores from newly generated tokens and updates only those tokens that require recomputation, while preserving image generation quality. MARch\'e significantly reduces redundant computation in MAR without modifying the underlying architecture. Empirically, MARch\'e achieves up to 1.7x speedup with negligible impact on image quality, offering a scalable and broadly applicable solution for efficient masked transformer generation.
Authors: Yanting Miao, William Loh, Suraj Kothawade, Pacal Poupart
Abstract: Recent work uses reinforcement learning (RL) to fine-tune text-to-image diffusion models, improving text-image alignment and sample quality. However, existing approaches introduce unnecessary complexity: they cache the full sampling trajectory, depend on differentiable reward models or large preference datasets, or require specialized guidance techniques. Motivated by the "golden noise" hypothesis -- that certain initial noise samples can consistently yield superior alignment -- we introduce Noise PPO, a minimalist RL algorithm that leaves the pre-trained diffusion model entirely frozen and learns a prompt-conditioned initial noise generator. Our approach requires no trajectory storage, reward backpropagation, or complex guidance tricks. Extensive experiments show that optimizing the initial noise distribution consistently improves alignment and sample quality over the original model, with the most significant gains at low inference steps. As the number of inference steps increases, the benefit of noise optimization diminishes but remains present. These findings clarify the scope and limitations of the golden noise hypothesis and reinforce the practical value of minimalist RL fine-tuning for diffusion models.
Authors: Zeyu Liu, Yunquan Zhang, Boyang Zhang, Guoyong Jiang, Daning Cheng
Abstract: Training large language models typically demands extensive GPU memory and substantial financial investment, which poses a barrier for many small- to medium-sized teams. In this paper, we present a full-parameter pre-training framework based on block coordinate descent (BCD), augmented with engineering optimizations, to efficiently train large models on affordable RTX 4090 GPU clusters. BCD ensures model convergence based on block coordinate descent theory and performs gradient computation and update at the level of parameter blocks. Experiments show that 1) Lower cost of Same-Device: BCD significantly reduces pre-training cost. For the 7B model, under identical hardware settings, BCD lowers training costs to approximately 33% on A100,A800 clusters on 7B model averagely and to approximately 2.6% on RTX 4090 clusters on 7B model, compared to traditional full-parameter training. 2) Cross-Device Transfer: By leveraging BCD, large-scale models previously trainable only on high-end A100 clusters can be seamlessly migrated and pre-trained on 4090 clusters-whose hourly cost is only one-quarter that of A100-without requiring expensive hardware. 3) Accuracy Retention: In both scenarios, BCD training achieves the same level of model accuracy as full-parameter pre-training.
Authors: Fangxin Liu, Ning Yang, Junping Zhao, Tao Yang, Haibing Guan, Li Jiang
Abstract: Large language models (LLMs) have achieved significant progress in natural language processing but face challenges in deployment due to high memory and computational requirements. Weight quantization is a common approach to address these issues, yet achieving effective low-bit compression remains challenging. This paper presents LCD, which unifies the learning of clustering-based quantization within a knowledge distillation framework. Using carefully designed optimization techniques, LCD preserves LLM performance even at ultra-low bit widths of 2-3 bits. Additionally, LCD compresses activations through smoothing and accelerates inference with a LUT-based design. Experimental results show that LCD outperforms existing methods and delivers up to a 6.2x speedup in inference. Notably, LCD is shown to be more cost-effective, making it a practical solution for real-world applications.
Authors: Leonardo Fonseca Larrubia, Pedro Alberto Morettin, Chang Chiann
Abstract: We present the Maximal Overlap Discrete Wavelet Scattering Transform (MODWST), whose construction is inspired by the combination of the Maximal Overlap Discrete Wavelet Transform (MODWT) and the Scattering Wavelet Transform (WST). We also discuss the use of MODWST in classification tasks, evaluating its performance in two applications: stationary signal classification and ECG signal classification. The results demonstrate that MODWST achieved good performance in both applications, positioning itself as a viable alternative to popular methods like Convolutional Neural Networks (CNNs), particularly when the training data set is limited.
Authors: Hao Gu, Lujun Li, Zheyu Wang, Bei Liu, Qiyuan Zhu, Sirui Han, Yike Guo
Abstract: Binary quantization represents the most extreme form of large language model (LLM) compression, reducing weights to $\pm$1 for maximal memory and computational efficiency. While recent sparsity-aware binarization methods achieve sub-1-bit compression by pruning redundant binary weights, they suffer from three critical challenges: performance deterioration, computational complexity from sparse mask management, and limited hardware compatibility. In this paper, we present BTC-LLM, a novel sub-1-bit LLM quantization framework that leverages adaptive weight transformation and binary pattern clustering to overcome these limitations, delivering both superior accuracy and efficiency. Our approach incorporates two key innovations: (1) a Learnable Transformation that optimizes invertible scaling and rotation matrices to align binarized weights with full-precision distributions, enabling incoherence processing to enhance layer-wise representation quality; (2) a Flash and Accurate Binary Codebook that identifies recurring binary vector clusters, compressing them into compact indices with tailored distance metrics and sign-based centroid updates. This eliminates the need for sparse masks, enabling efficient inference on standard hardware. Our code is available at https://github.com/Chooovy/BTC-LLM.
Authors: Yewei Liu, Xiyuan Wang, Muhan Zhang
Abstract: Network pruning, aimed at reducing network size while preserving accuracy, has attracted significant research interest. Numerous pruning techniques have been proposed over time. They are becoming increasingly effective, but more complex and harder to interpret as well. Given the inherent complexity of neural networks, we argue that manually designing pruning criteria has reached a bottleneck. To address this, we propose a novel approach in which we "use a neural network to prune neural networks". More specifically, we introduce the newly developed idea of metanetwork from meta-learning into pruning. A metanetwork is a network that takes another network as input and produces a modified network as output. In this paper, we first establish a bijective mapping between neural networks and graphs, and then employ a graph neural network as our metanetwork. We train a metanetwork that learns the pruning strategy automatically which can transform a network that is hard to prune into another network that is much easier to prune. Once the metanetwork is trained, our pruning needs nothing more than a feedforward through the metanetwork and the standard finetuning to prune at state-of-the-art. Our method achieved outstanding results on many popular and representative pruning tasks (including ResNet56 on CIFAR10, VGG19 on CIFAR100, ResNet50 on ImageNet). Our code is available at https://github.com/Yewei-Liu/MetaPruning
Authors: Alejandro Kuratomi, Zed Lee, Guilherme Dinis Chaliane Junior, Tony Lindgren, Diego Garc\'ia P\'erez
Abstract: Several interpretability methods for convolutional network-based classifiers exist. Most of these methods focus on extracting saliency maps for a given sample, providing a local explanation that highlights the main regions for the classification. However, some of these methods lack detailed explanations in the input space due to upscaling issues or may require random perturbations to extract the explanations. We propose Convolutional Rectifier for Interpretable Time Series Classification, or CRITS, as an interpretable model for time series classification that is designed to intrinsically extract local explanations. The proposed method uses a layer of convolutional kernels, a max-pooling layer and a fully-connected rectifier network (a network with only rectified linear unit activations). The rectified linear unit activation allows the extraction of the feature weights for the given sample, eliminating the need to calculate gradients, use random perturbations and the upscale of the saliency maps to the initial input space. We evaluate CRITS on a set of datasets, and study its classification performance and its explanation alignment, sensitivity and understandability.
Authors: Ting-Yun Chang, Muru Zhang, Jesse Thomason, Robin Jia
Abstract: Low-bit weight-only quantization significantly reduces the memory footprint of large language models (LLMs), but disproportionately affects certain examples. We analyze diverse 3-4 bit methods on LLMs ranging from 7B-70B in size and find that the quantization errors of 50 pairs of methods are strongly correlated (avg. 0.82) on FineWeb examples. Moreover, the residual stream magnitudes of full-precision models are indicative of future quantization errors. We further establish a hypothesis that relates the residual stream magnitudes to error amplification and accumulation over layers. Using LLM localization techniques, early exiting, and activation patching, we show that examples with large errors rely on precise residual activations in the late layers, and that the outputs of MLP gates play a crucial role in maintaining the perplexity. Our work reveals why certain examples result in large quantization errors and which model components are most critical for performance preservation.
Authors: Kazuma Kobayashi, Samrendra Roy, Seid Koric, Diab Abueidda, Syed Bahauddin Alam
Abstract: Accurate reconstruction of latent environmental fields from sparse and indirect observations is a foundational challenge across scientific domains-from atmospheric science and geophysics to public health and aerospace safety. Traditional approaches rely on physics-based simulators or dense sensor networks, both constrained by high computational cost, latency, or limited spatial coverage. We present the Temporal Radiation Operator Network (TRON), a spatiotemporal neural operator architecture designed to infer continuous global scalar fields from sequences of sparse, non-uniform proxy measurements. Unlike recent forecasting models that operate on dense, gridded inputs to predict future states, TRON addresses a more ill-posed inverse problem: reconstructing the current global field from sparse, temporally evolving sensor sequences, without access to future observations or dense labels. Demonstrated on global cosmic radiation dose reconstruction, TRON is trained on 22 years of simulation data and generalizes across 65,341 spatial locations, 8,400 days, and sequence lengths from 7 to 90 days. It achieves sub-second inference with relative L2 errors below 0.1%, representing a >58,000X speedup over Monte Carlo-based estimators. Though evaluated in the context of cosmic radiation, TRON offers a domain-agnostic framework for scientific field reconstruction from sparse data, with applications in atmospheric modeling, geophysical hazard monitoring, and real-time environmental risk forecasting.
Authors: Di Wu, Linghao Bu, Yifei Jia, Lu Cao, Siyuan Li, Siyu Chen, Yueqian Zhou, Sheng Fan, Wenjie Ren, Dengchang Wu, Kang Wang, Yue Zhang, Yuehui Ma, Jie Yang, Mohamad Sawan
Abstract: Recent achievements in implantable brain-computer interfaces (iBCIs) have demonstrated the potential to decode cognitive and motor behaviors with intracranial brain recordings; however, individual physiological and electrode implantation heterogeneities have constrained current approaches to neural decoding within single individuals, rendering interindividual neural decoding elusive. Here, we present Multi-individual Brain Region-Aggregated Network (MIBRAIN), a neural decoding framework that constructs a whole functional brain network model by integrating intracranial neurophysiological recordings across multiple individuals. MIBRAIN leverages self-supervised learning to derive generalized neural prototypes and supports group-level analysis of brain-region interactions and inter-subject neural synchrony. To validate our framework, we recorded stereoelectroencephalography (sEEG) signals from a cohort of individuals performing Mandarin syllable articulation. Both real-time online and offline decoding experiments demonstrated significant improvements in both audible and silent articulation decoding, enhanced decoding accuracy with increased multi-subject data integration, and effective generalization to unseen subjects. Furthermore, neural predictions for regions without direct electrode coverage were validated against authentic neural data. Overall, this framework paves the way for robust neural decoding across individuals and offers insights for practical clinical applications.
Authors: Jiajun He, Naoki Sawada, Koichi Miyazaki, Tomoki Toda
Abstract: In real-world applications, automatic speech recognition (ASR) systems must handle overlapping speech from multiple speakers and recognize rare words like technical terms. Traditional methods address multi-talker ASR and contextual biasing separately, limiting performance in complex scenarios. We propose a unified framework that combines multi-talker overlapping speech recognition and contextual biasing into a single task. Our ASR method integrates pretrained speech encoders and large language models (LLMs), using optimized finetuning strategies. We also introduce a two-stage filtering algorithm to efficiently identify relevant rare words from large biasing lists and incorporate them into the LLM's prompt input, enhancing rare word recognition. Experiments show that our approach outperforms traditional contextual biasing methods, achieving a WER of 7.9% on LibriMix and 32.9% on AMI SDM when the biasing size is 1,000, demonstrating its effectiveness in complex speech scenarios.
Authors: Christopher Nott
Abstract: Cybersecurity organizations are adapting to GenAI integration through modified frameworks and hybrid operational processes, with success influenced by existing security maturity, regulatory requirements, and investments in human capital and infrastructure. This qualitative research employs systematic document analysis and comparative case study methodology to examine how cybersecurity organizations adapt their threat modeling frameworks and operational processes to address generative artificial intelligence integration. Through examination of 25 studies from 2022 to 2025, the research documents substantial transformation in organizational approaches to threat modeling, moving from traditional signature-based systems toward frameworks incorporating artificial intelligence capabilities. The research identifies three primary adaptation patterns: Large Language Model integration for security applications, GenAI frameworks for risk detection and response automation, and AI/ML integration for threat hunting. Organizations with mature security infrastructures, particularly in finance and critical infrastructure sectors, demonstrate higher readiness through structured governance approaches, dedicated AI teams, and robust incident response processes. Organizations achieve successful GenAI integration when they maintain appropriate human oversight of automated systems, address data quality concerns and explainability requirements, and establish governance frameworks tailored to their specific sectors. Organizations encounter ongoing difficulties with privacy protection, bias reduction, personnel training, and defending against adversarial attacks. This work advances understanding of how organizations adopt innovative technologies in high-stakes environments and offers actionable insights for cybersecurity professionals implementing GenAI systems.
Authors: Aditya Kamlesh Parikh, Cristian Tejedor-Garcia, Catia Cucchiarini, Helmer Strik
Abstract: Pronunciation assessment relies on goodness of pronunciation (GOP) scores, traditionally derived from softmax-based posterior probabilities. However, posterior probabilities may suffer from overconfidence and poor phoneme separation, limiting their effectiveness. This study compares logit-based GOP scores with probability-based GOP scores for mispronunciation detection. We conducted our experiment on two L2 English speech datasets spoken by Dutch and Mandarin speakers, assessing classification performance and correlation with human ratings. Logit-based methods outperform probability-based GOP in classification, but their effectiveness depends on dataset characteristics. The maximum logit GOP shows the strongest alignment with human perception, while a combination of different GOP scores balances probability and logit features. The findings suggest that hybrid GOP methods incorporating uncertainty modeling and phoneme-specific weighting improve pronunciation assessment.
Authors: Joydeep Chandra, Aleksandr Algazinov, Satyam Kumar Navneet, Rim El Filali, Matt Laing, Andrew Hanna
Abstract: As access to information becomes more open and widespread, people are increasingly using AI tools for assistance. However, many of these tools struggle to estimate the trustworthiness of the information. Although today's search engines include AI features, they often fail to offer clear indicators of data reliability. To address this gap, we introduce WebTrust, a system designed to simplify the process of finding and judging credible information online. Built on a fine-tuned version of IBM's Granite-1B model and trained on a custom dataset, WebTrust works by assigning a reliability score (from 0.1 to 1) to each statement it processes. In addition, it offers a clear justification for why a piece of information received that score. Evaluated using prompt engineering, WebTrust consistently achieves superior performance compared to other small-scale LLMs and rule-based approaches, outperforming them across all experiments on MAE, RMSE, and R2. User testing showed that when reliability scores are displayed alongside search results, people feel more confident and satisfied with the information they find. With its accuracy, transparency, and ease of use, WebTrust offers a practical solution to help combat misinformation and make trustworthy information more accessible to everyone.
Authors: Zongli Ye, Jiachen Lian, Xuanru Zhou, Jinming Zhang, Haodong Li, Shuhe Li, Chenxu Guo, Anaisha Das, Peter Park, Zoe Ezzes, Jet Vonk, Brittany Morin, Rian Bogley, Lisa Wauters, Zachary Miller, Maria Gorno-Tempini, Gopala Anumanchipalli
Abstract: Accurate alignment of dysfluent speech with intended text is crucial for automating the diagnosis of neurodegenerative speech disorders. Traditional methods often fail to model phoneme similarities effectively, limiting their performance. In this work, we propose Neural LCS, a novel approach for dysfluent text-text and speech-text alignment. Neural LCS addresses key challenges, including partial alignment and context-aware similarity mapping, by leveraging robust phoneme-level modeling. We evaluate our method on a large-scale simulated dataset, generated using advanced data simulation techniques, and real PPA data. Neural LCS significantly outperforms state-of-the-art models in both alignment accuracy and dysfluent speech segmentation. Our results demonstrate the potential of Neural LCS to enhance automated systems for diagnosing and analyzing speech disorders, offering a more accurate and linguistically grounded solution for dysfluent speech alignment.
Authors: Nirmal Gelal, Chloe Snow, Ambyr Rios, Hande K\"u\c{c}\"uk McGinty
Abstract: The implementation of transformational pedagogy in secondary education classrooms requires a broad multiliteracy approach. Due to limited planning time and resources, high school English Literature teachers often struggle to curate diverse, thematically aligned literature text sets. This study addresses the critical need for a tool that provides scaffolds for novice educators in selecting literature texts that are diverse -- in terms of genre, theme, subtheme, and author -- yet similar in context and pedagogical merits. We have developed a recommendation system, Teaching Text Expansion for Teacher Scaffolding (T-TExTS), that suggests high school English Literature books based on pedagogical merits, genre, and thematic relevance using a knowledge graph. We constructed a domain-specific ontology using the KNowledge Acquisition and Representation Methodology (KNARM), transformed into a knowledge graph, which was then embedded using DeepWalk, biased random walk, and a hybrid of both approaches. The system was evaluated using link prediction and recommendation performance metrics, including Area Under the Curve (AUC), Mean Reciprocal Rank (MRR), Hits@K, and normalized Discounted Cumulative Gain (nDCG). DeepWalk outperformed in most ranking metrics, with the highest AUC (0.9431), whereas the hybrid model offered balanced performance. These findings demonstrate the importance of semantic, ontology-driven approaches in recommendation systems and suggest that T-TExTS can significantly ease the burden of English Literature text selection for high school educators, promoting more informed and inclusive curricular decisions. The source code for T-TExTS is available at: https://github.com/koncordantlab/TTExTS
Authors: Assaf Marron
Abstract: We present a handcrafted neural network that, without training, solves the seemingly difficult problem of encoding an arbitrary set of integers into a single numerical variable, and then recovering the original elements. While using only standard neural network operations -- weighted sums with biases and identity activation -- we make design choices that challenge common notions in this area around representation, continuity of domains, computation, learnability and more. For example, our construction is designed, not learned; it represents multiple values using a single one by simply concatenating digits without compression, and it relies on hardware-level truncation of rightmost digits as a bit-manipulation mechanism. This neural net is not intended for practical application. Instead, we see its resemblance to -- and deviation from -- standard trained autoencoders as an invitation to examine assumptions that may unnecessarily constrain the development of systems and models based on autoencoding and machine learning. Motivated in part by our research on a theory of biological evolution centered around natural autoencoding of species characteristics, we conclude by refining the discussion with a biological perspective.
Authors: Haoxiang Guan, Jiyan He, Liyang Fan, Zhenzhen Ren, Shaobin He, Xin Yu, Yuan Chen, Shuxin Zheng, Tie-Yan Liu, Zhen Liu
Abstract: Understanding how complex societal behaviors emerge from individual cognition and interactions requires both high-fidelity modeling of human behavior and large-scale simulations. Traditional agent-based models (ABMs) have been employed to study these dynamics for decades, but are constrained by simplified agent behaviors that fail to capture human complexity. Recent advances in large language models (LLMs) offer new opportunities by enabling agents to exhibit sophisticated social behaviors that go beyond rule-based logic, yet face significant scaling challenges. Here we present Light Society, an agent-based simulation framework that advances both fronts, efficiently modeling human-like societies at planetary scale powered by LLMs. Light Society formalizes social processes as structured transitions of agent and environment states, governed by a set of LLM-powered simulation operations, and executed through an event queue. This modular design supports both independent and joint component optimization, supporting efficient simulation of societies with over one billion agents. Large-scale simulations of trust games and opinion propagation--spanning up to one billion agents--demonstrate Light Society's high fidelity and efficiency in modeling social trust and information diffusion, while revealing scaling laws whereby larger simulations yield more stable and realistic emergent behaviors.
Authors: Shaba Shaon, Van-Dinh Nguyen, Dinh C. Nguyen
Abstract: In this paper, we study a novel latency minimization problem in wireless federated learning (FL) across multi-hop networks. The system comprises multiple routes, each integrating leaf and relay nodes for FL model training. We explore a personalized learning and adaptive aggregation-aware FL (PAFL) framework that effectively addresses data heterogeneity across participating nodes by harmonizing individual and collective learning objectives. We formulate an optimization problem aimed at minimizing system latency through the joint optimization of leaf and relay nodes, as well as relay routing indicator. We also incorporate an additional energy harvesting scheme for the relay nodes to help with their relay tasks. This formulation presents a computationally demanding challenge, and thus we develop a simple yet efficient algorithm based on block coordinate descent and successive convex approximation (SCA) techniques. Simulation results illustrate the efficacy of our proposed joint optimization approach for leaf and relay nodes with relay routing indicator. We observe significant latency savings in the wireless multi-hop PAFL system, with reductions of up to 69.37% compared to schemes optimizing only one node type, traditional greedy algorithm, and scheme without relay routing indicator.
Authors: Michele Alberti (LSL), Fran\c{c}ois Bobot (LSL), Julien Girard-Satabin (LSL), Alban Grastien (LSL), Aymeric Varasse (LSL), Zakaria Chihani (LSL)
Abstract: The formal specification and verification of machine learning programs saw remarkable progress in less than a decade, leading to a profusion of tools. However, diversity may lead to fragmentation, resulting in tools that are difficult to compare, except for very specific benchmarks. Furthermore, this progress is heavily geared towards the specification and verification of a certain class of property, that is, local robustness properties. But while provers are becoming more and more efficient at solving local robustness properties, even slightly more complex properties, involving multiple neural networks for example, cannot be expressed in the input languages of winners of the International Competition of Verification of Neural Networks VNN-Comp. In this tool paper, we present CAISAR, an open-source platform dedicated to machine learning specification and verification. We present its specification language, suitable for modelling complex properties on neural networks, support vector machines and boosted trees. We show on concrete use-cases how specifications written in this language are automatically translated to queries to state-of-the-art provers, notably by using automated graph editing techniques, making it possible to use their off-the-shelf versions. The artifact to reproduce the paper claims is available at the following DOI: https://doi.org/10.5281/zenodo.15209510
Authors: Chrisantha Fernando, Dylan Banarse, Simon Osindero
Abstract: Because we are highly motivated to be understood, we created public external representations -- mime, language, art -- to externalise our inner states. We argue that such external representations are a pre-condition for access consciousness, the global availability of information for reasoning. Yet the bandwidth of access consciousness is tiny compared with the richness of `raw experience', so no external representation can reproduce that richness in full. Ordinarily an explanation of experience need only let an audience `grasp' the relevant pattern, not relive the phenomenon. But our drive to be understood, and our low level sensorimotor capacities for `grasping' so rich, that the demand for an explanation of the feel of experience cannot be ``satisfactory''. That inflated epistemic demand (the preeminence of our expectation that we could be perfectly understood by another or ourselves) rather than an irreducible metaphysical gulf -- keeps the hard problem of consciousness alive. But on the plus side, it seems we will simply never give up creating new ways to communicate and think about our experiences. In this view, to be consciously aware is to strive to have one's agency understood by oneself and others.
Authors: Wanjin Feng, Xingyu Gao, Wenqian Du, Hailong Shi, Peilin Zhao, Pengcheng Wu, Chunyan Miao
Abstract: Spiking Neural Networks (SNNs) often suffer from high time complexity $O(T)$ due to the sequential processing of $T$ spikes, making training computationally expensive. In this paper, we propose a novel Fixed-point Parallel Training (FPT) method to accelerate SNN training without modifying the network architecture or introducing additional assumptions. FPT reduces the time complexity to $O(K)$, where $K$ is a small constant (usually $K=3$), by using a fixed-point iteration form of Leaky Integrate-and-Fire (LIF) neurons for all $T$ timesteps. We provide a theoretical convergence analysis of FPT and demonstrate that existing parallel spiking neurons can be viewed as special cases of our proposed method. Experimental results show that FPT effectively simulates the dynamics of original LIF neurons, significantly reducing computational time without sacrificing accuracy. This makes FPT a scalable and efficient solution for real-world applications, particularly for long-term tasks. Our code will be released at \href{https://github.com/WanjinVon/FPT}{\texttt{https://github.com/WanjinVon/FPT}}.
URLs: https://github.com/WanjinVon/FPT, https://github.com/WanjinVon/FPT
Authors: Muhammad Sukri Bin Ramli
Abstract: Tariff exemptions are fundamental to attracting Foreign Direct Investment (FDI) into the manufacturing sector, though the associated administrative processes present areas for optimization for both investing entities and the national tax authority. This paper proposes a conceptual framework to empower tax administration by leveraging a synergistic integration of Optical Character Recognition (OCR) and Large Language Model (LLM) technologies. The proposed system is designed to first utilize OCR for intelligent digitization, precisely extracting data from diverse application documents and key regulatory texts such as tariff orders. Subsequently, the LLM would enhance the capabilities of administrative officers by automating the critical and time-intensive task of verifying submitted HS Tariff Codes for machinery, equipment, and raw materials against official exemption lists. By enhancing the speed and precision of these initial assessments, this AI-driven approach systematically reduces potential for non-alignment and non-optimized exemption utilization, thereby streamlining the investment journey for FDI companies. For the national administration, the benefits include a significant boost in operational capacity, reduced administrative load, and a strengthened control environment, ultimately improving the ease of doing business and solidifying the nation's appeal as a premier destination for high-value manufacturing FDI.
Authors: Timothy Dubber, Seth Lazar
Abstract: This paper argues that autonomous AI cyber-weapons - Military-AI Cyber Agents (MAICAs) - create a credible pathway to catastrophic risk. It sets out the technical feasibility of MAICAs, explains why geopolitics and the nature of cyberspace make MAICAs a catastrophic risk, and proposes political, defensive-AI and analogue-resilience measures to blunt the threat.
Authors: Naba Rizvi, Taggert Smith, Tanvi Vidyala, Mya Bolds, Harper Strickland, Andrew Begel, Rua Williams, Imani Munyaka
Abstract: Human-like AI agents such as robots and chatbots are becoming increasingly popular, but they present a variety of ethical concerns. The first concern is in how we define humanness, and how our definition impacts communities historically dehumanized by scientific research. Autistic people in particular have been dehumanized by being compared to robots, making it even more important to ensure this marginalization is not reproduced by AI that may promote neuronormative social behaviors. Second, the ubiquitous use of these agents raises concerns surrounding model biases and accessibility. In our work, we investigate the experiences of the people who build and design these technologies to gain insights into their understanding and acceptance of neurodivergence, and the challenges in making their work more accessible to users with diverse needs. Even though neurodivergent individuals are often marginalized for their unique communication styles, nearly all participants overlooked the conclusions their end-users and other AI system makers may draw about communication norms from the implementation and interpretation of humanness applied in participants' work. This highlights a major gap in their broader ethical considerations, compounded by some participants' neuronormative assumptions about the behaviors and traits that distinguish "humans" from "bots" and the replication of these assumptions in their work. We examine the impact this may have on autism inclusion in society and provide recommendations for additional systemic changes towards more ethical research directions.
Authors: Thabassum Aslam, Anees Aslam
Abstract: SocialCredit+ is AI powered credit scoring system that leverages publicly available social media data to augment traditional credit evaluation. It uses a conversational banking assistant to gather user consent and fetch public profiles. Multimodal feature extractors analyze posts, bios, images, and friend networks to generate a rich behavioral profile. A specialized Sharia-compliance layer flags any non-halal indicators and prohibited financial behavior based on Islamic ethics. The platform employs a retrieval-augmented generation module: an LLM accesses a domain specific knowledge base to generate clear, text-based explanations for each decision. We describe the end-to-end architecture and data flow, the models used, and system infrastructure. Synthetic scenarios illustrate how social signals translate into credit-score factors. This paper emphasizes conceptual novelty, compliance mechanisms, and practical impact, targeting AI researchers, fintech practitioners, ethical banking jurists, and investors.
Authors: Reza Fayyazi, Michael Zuzak, Shanchieh Jay Yang
Abstract: Security vulnerabilities are rapidly increasing in frequency and complexity, creating a shifting threat landscape that challenges cybersecurity defenses. Large Language Models (LLMs) have been widely adopted for cybersecurity threat analysis. When querying LLMs, dealing with new, unseen vulnerabilities is particularly challenging as it lies outside LLMs' pre-trained distribution. Retrieval-Augmented Generation (RAG) pipelines mitigate the problem by injecting up-to-date authoritative sources into the model context, thus reducing hallucinations and increasing the accuracy in responses. Meanwhile, the deployment of LLMs in security-sensitive environments introduces challenges around trust and safety. This raises a critical open question: How to quantify or attribute the generated response to the retrieved context versus the model's pre-trained knowledge? This work proposes LLM Embedding-based Attribution (LEA) -- a novel, explainable metric to paint a clear picture on the 'percentage of influence' the pre-trained knowledge vs. retrieved content has for each generated response. We apply LEA to assess responses to 100 critical CVEs from the past decade, verifying its effectiveness to quantify the insightfulness for vulnerability analysis. Our development of LEA reveals a progression of independency in hidden states of LLMs: heavy reliance on context in early layers, which enables the derivation of LEA; increased independency in later layers, which sheds light on why scale is essential for LLM's effectiveness. This work provides security analysts a means to audit LLM-assisted workflows, laying the groundwork for transparent, high-assurance deployments of RAG-enhanced LLMs in cybersecurity operations.
Authors: Hao Li, Xiaogeng Liu, Hung-Chun Chiu, Dianqi Li, Ning Zhang, Chaowei Xiao
Abstract: Large Language Models (LLMs) are increasingly central to agentic systems due to their strong reasoning and planning capabilities. By interacting with external environments through predefined tools, these agents can carry out complex user tasks. Nonetheless, this interaction also introduces the risk of prompt injection attacks, where malicious inputs from external sources can mislead the agent's behavior, potentially resulting in economic loss, privacy leakage, or system compromise. System-level defenses have recently shown promise by enforcing static or predefined policies, but they still face two key challenges: the ability to dynamically update security rules and the need for memory stream isolation. To address these challenges, we propose DRIFT, a Dynamic Rule-based Isolation Framework for Trustworthy agentic systems, which enforces both control- and data-level constraints. A Secure Planner first constructs a minimal function trajectory and a JSON-schema-style parameter checklist for each function node based on the user query. A Dynamic Validator then monitors deviations from the original plan, assessing whether changes comply with privilege limitations and the user's intent. Finally, an Injection Isolator detects and masks any instructions that may conflict with the user query from the memory stream to mitigate long-term risks. We empirically validate the effectiveness of DRIFT on the AgentDojo benchmark, demonstrating its strong security performance while maintaining high utility across diverse models -- showcasing both its robustness and adaptability.
Authors: Bassam Noori Shaker, Bahaa Al-Musawi, Mohammed Falih Hassan
Abstract: An Advanced Persistent Threat (APT) is a multistage, highly sophisticated, and covert form of cyber threat that gains unauthorized access to networks to either steal valuable data or disrupt the targeted network. These threats often remain undetected for extended periods, emphasizing the critical need for early detection in networks to mitigate potential APT consequences. In this work, we propose a feature selection method for developing a lightweight intrusion detection system capable of effectively identifying APTs at the initial compromise stage. Our approach leverages the XGBoost algorithm and Explainable Artificial Intelligence (XAI), specifically utilizing the SHAP (SHapley Additive exPlanations) method for identifying the most relevant features of the initial compromise stage. The results of our proposed method showed the ability to reduce the selected features of the SCVIC-APT-2021 dataset from 77 to just four while maintaining consistent evaluation metrics for the suggested system. The estimated metrics values are 97% precision, 100% recall, and a 98% F1 score. The proposed method not only aids in preventing successful APT consequences but also enhances understanding of APT behavior at early stages.
Authors: Hyungjune Bu, Chanjoo Jung, Minjae Kang, Jaehyung Kim
Abstract: As large language models (LLMs) are progressively deployed in various real-world applications, personalization of LLMs has become increasingly important. While various approaches to LLM personalization such as prompt-based and training-based methods have been actively explored, the development of effective decoding-time algorithms remains largely overlooked, despite their demonstrated potential. In this paper, we propose CoPe (Contrasting Personal Preference), a novel decoding-time approach applied after performing parameter-efficient fine-tuning (PEFT) on user-specific data. Our core idea is to leverage reward-guided decoding specifically for personalization by maximizing each user's implicit reward signal. We evaluate CoPe across five open-ended personalized text generation tasks. Our empirical results demonstrate that CoPe achieves strong performance, improving personalization by an average of 10.57% in ROUGE-L, without relying on external reward models or additional training procedures.
Authors: Qirui Mi, Qipeng Yang, Zijun Fan, Wentian Fan, Heyang Ma, Chengdong Ma, Siyu Xia, Bo An, Jun Wang, Haifeng Zhang
Abstract: Artificial intelligence (AI) has become a powerful tool for economic research, enabling large-scale simulation and policy optimization. However, applying AI effectively requires simulation platforms for scalable training and evaluation-yet existing environments remain limited to simplified, narrowly scoped tasks, falling short of capturing complex economic challenges such as demographic shifts, multi-government coordination, and large-scale agent interactions. To address this gap, we introduce EconGym, a scalable and modular testbed that connects diverse economic tasks with AI algorithms. Grounded in rigorous economic modeling, EconGym implements 11 heterogeneous role types (e.g., households, firms, banks, governments), their interaction mechanisms, and agent models with well-defined observations, actions, and rewards. Users can flexibly compose economic roles with diverse agent algorithms to simulate rich multi-agent trajectories across 25+ economic tasks for AI-driven policy learning and analysis. Experiments show that EconGym supports diverse and cross-domain tasks-such as coordinating fiscal, pension, and monetary policies-and enables benchmarking across AI, economic methods, and hybrids. Results indicate that richer task composition and algorithm diversity expand the policy space, while AI agents guided by classical economic methods perform best in complex settings. EconGym also scales to 10k agents with high realism and efficiency.
Authors: Oscar Boullosa Dapena
Abstract: Real-time continuous learning over streaming data remains a central challenge in deep learning and AI systems. Traditional gradient-based models such as backpropagation through time (BPTT) face computational and stability limitations when dealing with temporally unbounded data. In this paper, we introduce a novel architecture, Quantum-Inspired Differentiable Integral Neural Networks (QIDINNs), which leverages the Feynman technique of differentiation under the integral sign to formulate neural updates as integrals over historical data. This reformulation allows for smoother, more stable learning dynamics that are both physically interpretable and computationally tractable. Inspired by Feynman's path integral formalism and compatible with quantum gradient estimation frameworks, QIDINNs open a path toward hybrid classical-quantum neural computation. We demonstrate our model's effectiveness on synthetic and real-world streaming tasks, and we propose directions for quantum extensions and scalable implementations.
Authors: Benjamin Marais, Tony Quertier, Gr\'egoire Barrue
Abstract: In a context of malware analysis, numerous approaches rely on Artificial Intelligence to handle a large volume of data. However, these techniques focus on data view (images, sequences) and not on an expert's view. Noticing this issue, we propose a preprocessing that focuses on expert knowledge to improve malware semantic analysis and result interpretability. We propose a new preprocessing method which creates JSON reports for Portable Executable files. These reports gather features from both static and behavioral analysis, and incorporate packer signature detection, MITRE ATT\&CK and Malware Behavior Catalog (MBC) knowledge. The purpose of this preprocessing is to gather a semantic representation of binary files, understandable by malware analysts, and that can enhance AI models' explainability for malicious files analysis. Using this preprocessing to train a Large Language Model for Malware classification, we achieve a weighted-average F1-score of 0.94 on a complex dataset, representative of market reality.
Authors: Brown Ebouky, Andrea Bartezzaghi, Mattia Rigotti
Abstract: The recent advent of reasoning models like OpenAI's o1 was met with excited speculation by the AI community about the mechanisms underlying these capabilities in closed models, followed by a rush of replication efforts, particularly from the open source community. These speculations were largely settled by the demonstration from DeepSeek-R1 that chains-of-thought and reinforcement learning (RL) can effectively replicate reasoning on top of base LLMs. However, it remains valuable to explore alternative methods for theoretically eliciting reasoning that could help elucidate the underlying mechanisms, as well as providing additional methods that may offer complementary benefits. Here, we build on the long-standing literature in cognitive psychology and cognitive architectures, which postulates that reasoning arises from the orchestrated, sequential execution of a set of modular, predetermined cognitive operations. Crucially, we implement this key idea within a modern agentic tool-calling framework. In particular, we endow an LLM with a small set of "cognitive tools" encapsulating specific reasoning operations, each executed by the LLM itself. Surprisingly, this simple strategy results in considerable gains in performance on standard mathematical reasoning benchmarks compared to base LLMs, for both closed and open-weight models. For instance, providing our "cognitive tools" to GPT-4.1 increases its pass@1 performance on AIME2024 from 26.7% to 43.3%, bringing it very close to the performance of o1-preview. In addition to its practical implications, this demonstration contributes to the debate regarding the role of post-training methods in eliciting reasoning in LLMs versus the role of inherent capabilities acquired during pre-training, and whether post-training merely uncovers these latent abilities.
Authors: Phillipe R. Sampaio, Helene Maxcici
Abstract: This paper investigates a novel approach to unsupervised document clustering by leveraging multimodal embeddings as input to traditional clustering algorithms such as $k$-Means and DBSCAN. Our method aims to achieve a finer-grained document understanding by not only grouping documents at the type level (e.g., invoices, purchase orders), but also distinguishing between different templates within the same document category. This is achieved by using embeddings that capture textual content, layout information, and visual features of documents. We evaluated the effectiveness of this approach using embeddings generated by several state-of-the-art pretrained multimodal models, including SBERT, LayoutLMv1, LayoutLMv3, DiT, Donut, and ColPali. Our findings demonstrate the potential of multimodal embeddings to significantly enhance document clustering, offering benefits for various applications in intelligent document processing, document layout analysis, and unsupervised document classification. This work provides valuable insight into the advantages and limitations of different multimodal models for this task and opens new avenues for future research to understand and organize document collections.
Authors: Junjie Yu, Wenxiao Ma, Jianyu Zhang, Haotian Deng, Zihan Deng, Yi Guo, Quanying Liu
Abstract: Despite variations in architecture and pretraining strategies, recent studies indicate that large-scale AI models often converge toward similar internal representations that also align with neural activity. We propose that scale-invariance, a fundamental structural principle in natural systems, is a key driver of this convergence. In this work, we propose a multi-scale analytical framework to quantify two core aspects of scale-invariance in AI representations: dimensional stability and structural similarity across scales. We further investigate whether these properties can predict alignment performance with functional Magnetic Resonance Imaging (fMRI) responses in the visual cortex. Our analysis reveals that embeddings with more consistent dimension and higher structural similarity across scales align better with fMRI data. Furthermore, we find that the manifold structure of fMRI data is more concentrated, with most features dissipating at smaller scales. Embeddings with similar scale patterns align more closely with fMRI data. We also show that larger pretraining datasets and the inclusion of language modalities enhance the scale-invariance properties of embeddings, further improving neural alignment. Our findings indicate that scale-invariance is a fundamental structural principle that bridges artificial and biological representations, providing a new framework for evaluating the structural quality of human-like AI systems.
Authors: Houyi Li, Ka Man Lo, Ziqi Wang, Zili Wang, Wenzhen Zheng, Shuigeng Zhou, Xiangyu Zhang, Daxin Jiang
Abstract: Mixture-of-Experts (MoE) language models dramatically expand model capacity and achieve remarkable performance without increasing per-token compute. However, can MoEs surpass dense architectures under strictly equal resource constraints - that is, when the total parameter count, training compute, and data budget are identical? This question remains under-explored despite its significant practical value and potential. In this paper, we propose a novel perspective and methodological framework to study this question thoroughly. First, we comprehensively investigate the architecture of MoEs and achieve an optimal model design that maximizes the performance. Based on this, we subsequently find that an MoE model with activation rate in an optimal region is able to outperform its dense counterpart under the same total parameter, training compute and data resource. More importantly, this optimal region remains consistent across different model sizes. Although additional amount of data turns out to be a trade-off for the enhanced performance, we show that this can be resolved via reusing data. We validate our findings through extensive experiments, training nearly 200 language models at 2B scale and over 50 at 7B scale, cumulatively processing 50 trillion tokens. All models will be released publicly.
Authors: Shehroz S. Khan, Ali Abedi, Charlene H. Chu
Abstract: Interpreting large volumes of high-dimensional, unlabeled data in a manner that is comprehensible to humans remains a significant challenge across various domains. In unsupervised healthcare data analysis, interpreting clustered data can offer meaningful insights into patients' health outcomes, which hold direct implications for healthcare providers. This paper addresses the problem of interpreting clustered sensor data collected from older adult patients recovering from lower-limb fractures in the community. A total of 560 days of multimodal sensor data, including acceleration, step count, ambient motion, GPS location, heart rate, and sleep, alongside clinical scores, were remotely collected from patients at home. Clustering was first carried out separately for each data modality to assess the impact of feature sets extracted from each modality on patients' recovery trajectories. Then, using context-aware prompting, a large language model was employed to infer meaningful cluster labels for the clusters derived from each modality. The quality of these clusters and their corresponding labels was validated through rigorous statistical testing and visualization against clinical scores collected alongside the multimodal sensor data. The results demonstrated the statistical significance of most modality-specific cluster labels generated by the large language model with respect to clinical scores, confirming the efficacy of the proposed method for interpreting sensor data in an unsupervised manner. This unsupervised data analysis approach, relying solely on sensor data, enables clinicians to identify at-risk patients and take timely measures to improve health outcomes.
Authors: Huanqiang Duan, Manno Versluis, Qinyu Chen, Leo C. N. de Vreede, Chang Gao
Abstract: Digital predistortion (DPD) is essential for mitigating nonlinearity in RF power amplifiers, particularly for wideband applications. This paper presents TCN-DPD, a parameter-efficient architecture based on temporal convolutional networks, integrating noncausal dilated convolutions with optimized activation functions. Evaluated on the OpenDPD framework with the DPA_200MHz dataset, TCN-DPD achieves simulated ACPRs of -51.58/-49.26 dBc (L/R), EVM of -47.52 dB, and NMSE of -44.61 dB with 500 parameters and maintains superior linearization than prior models down to 200 parameters, making it promising for efficient wideband PA linearization.
Authors: Haoyu Dong, Yuwen Chen, Hanxue Gu, Nicholas Konz, Yaqian Chen, Qihang Li, Maciej A. Mazurowski
Abstract: The widespread use of Magnetic Resonance Imaging (MRI) and the rise of deep learning have enabled the development of powerful predictive models for a wide range of diagnostic tasks in MRI, such as image classification or object segmentation. However, training models for specific new tasks often requires large amounts of labeled data, which is difficult to obtain due to high annotation costs and data privacy concerns. To circumvent this issue, we introduce MRI-CORE (MRI COmprehensive Representation Encoder), a vision foundation model pre-trained using more than 6 million slices from over 110,000 MRI volumes across 18 main body locations. Experiments on five diverse object segmentation tasks in MRI demonstrate that MRI-CORE can significantly improve segmentation performance in realistic scenarios with limited labeled data availability, achieving an average gain of 6.97% 3D Dice Coefficient using only 10 annotated slices per task. We further demonstrate new model capabilities in MRI such as classification of image properties including body location, sequence type and institution, and zero-shot segmentation. These results highlight the value of MRI-CORE as a generalist vision foundation model for MRI, potentially lowering the data annotation resource barriers for many applications.
Authors: Pranav Agarwal, Ioana Ciuc\u{a}
Abstract: Large Language Models (LLMs) are increasingly integrated into everyday applications. As their influence grows, understanding their decision making and underlying personality becomes essential. In this work, we interpret model personality using our proposed Supernova Event Dataset, a novel dataset with diverse articles spanning biographies, historical events, news, and scientific discoveries. We use this dataset to benchmark LLMs on extracting and ranking key events from text, a subjective and complex challenge that requires reasoning over long-range context and modeling causal chains. We evaluate small models like Phi-4, Orca 2, and Qwen 2.5, and large, stronger models such as Claude 3.7, Gemini 2.5, and OpenAI o3, and propose a framework where another LLM acts as a judge to infer each model's personality based on its selection and classification of events. Our analysis shows distinct personality traits: for instance, Orca 2 demonstrates emotional reasoning focusing on interpersonal dynamics, while Qwen 2.5 displays a more strategic, analytical style. When analyzing scientific discovery events, Claude Sonnet 3.7 emphasizes conceptual framing, Gemini 2.5 Pro prioritizes empirical validation, and o3 favors step-by-step causal reasoning. This analysis improves model interpretability, making them user-friendly for a wide range of diverse applications.
Authors: Naomi Fridman, Bubby Solway, Tomer Fridman, Itamar Barnea, Anat Goldshtein
Abstract: Breast cancer remains a leading cause of cancer-related mortality worldwide, making early detection and accurate treatment response monitoring critical priorities. We present BreastDCEDL, a curated, deep learning-ready dataset comprising pre-treatment 3D Dynamic Contrast-Enhanced MRI (DCE-MRI) scans from 2,070 breast cancer patients drawn from the I-SPY1, I-SPY2, and Duke cohorts, all sourced from The Cancer Imaging Archive. The raw DICOM imaging data were rigorously converted into standardized 3D NIfTI volumes with preserved signal integrity, accompanied by unified tumor annotations and harmonized clinical metadata including pathologic complete response (pCR), hormone receptor (HR), and HER2 status. Although DCE-MRI provides essential diagnostic information and deep learning offers tremendous potential for analyzing such complex data, progress has been limited by lack of accessible, public, multicenter datasets. BreastDCEDL addresses this gap by enabling development of advanced models, including state-of-the-art transformer architectures that require substantial training data. To demonstrate its capacity for robust modeling, we developed the first transformer-based model for breast DCE-MRI, leveraging Vision Transformer (ViT) architecture trained on RGB-fused images from three contrast phases (pre-contrast, early post-contrast, and late post-contrast). Our ViT model achieved state-of-the-art pCR prediction performance in HR+/HER2- patients (AUC 0.94, accuracy 0.93). BreastDCEDL includes predefined benchmark splits, offering a framework for reproducible research and enabling clinically meaningful modeling in breast cancer imaging.
Authors: Jaeyeon Kim, Heeseung Yun, Gunhee Kim
Abstract: Spatial audio is essential for enhancing the immersiveness of audio-visual experiences, yet its production typically demands complex recording systems and specialized expertise. In this work, we address a novel problem of generating first-order ambisonics, a widely used spatial audio format, directly from silent videos. To support this task, we introduce YT-Ambigen, a dataset comprising 102K 5-second YouTube video clips paired with corresponding first-order ambisonics. We also propose new evaluation metrics to assess the spatial aspect of generated audio based on audio energy maps and saliency metrics. Furthermore, we present Video-to-Spatial Audio Generation (ViSAGe), an end-to-end framework that generates first-order ambisonics from silent video frames by leveraging CLIP visual features, autoregressive neural audio codec modeling with both directional and visual guidance. Experimental results demonstrate that ViSAGe produces plausible and coherent first-order ambisonics, outperforming two-stage approaches consisting of video-to-audio generation and audio spatialization. Qualitative examples further illustrate that ViSAGe generates temporally aligned high-quality spatial audio that adapts to viewpoint changes.
Authors: Stephen Mell, Botong Zhang, David Mell, Shuo Li, Ramya Ramalingam, Nathan Yu, Steve Zdancewic, Osbert Bastani
Abstract: Modern large language models (LLMs) are often deployed as agents, calling external tools adaptively to solve tasks. Rather than directly calling tools, it can be more effective for LLMs to write code to perform the tool calls, enabling them to automatically generate complex control flow such as conditionals and loops. Such code actions are typically provided as Python code, since LLMs are quite proficient at it; however, Python may not be the ideal language due to limited built-in support for performance, security, and reliability. We propose a novel programming language for code actions, called Quasar, which has several benefits: (1) automated parallelization to improve performance, (2) uncertainty quantification to improve reliability and mitigate hallucinations, and (3) security features enabling the user to validate actions. LLMs can write code in a subset of Python, which is automatically transpiled to Quasar. We evaluate our approach on the ViperGPT visual question answering agent, applied to the GQA dataset, demonstrating that LLMs with Quasar actions instead of Python actions retain strong performance, while reducing execution time when possible by 42%, improving security by reducing user approval interactions when possible by 52%, and improving reliability by applying conformal prediction to achieve a desired target coverage level.
Authors: Wenyue Hua, Dujian Ding, Yile Gu, Yujie Ren, Kai Mei, Minghua Ma, William Yang Wang
Abstract: Conventional operating system scheduling algorithms are largely content-ignorant, making decisions based on factors such as latency or fairness without considering the actual intents or semantics of processes. Consequently, these algorithms often do not prioritize tasks that require urgent attention or carry higher importance, such as in emergency management scenarios. However, recent advances in language models enable semantic analysis of processes, allowing for more intelligent and context-aware scheduling decisions. In this paper, we introduce the concept of semantic scheduling in scheduling of requests from large language models (LLM), where the semantics of the process guide the scheduling priorities. We present a novel scheduling algorithm with optimal time complexity, designed to minimize the overall waiting time in LLM-based prompt scheduling. To illustrate its effectiveness, we present a medical emergency management application, underscoring the potential benefits of semantic scheduling for critical, time-sensitive tasks. The code and data are available at https://github.com/Wenyueh/latency_optimization_with_priority_constraints.
URLs: https://github.com/Wenyueh/latency_optimization_with_priority_constraints.
Authors: Xudong Zhu, Jiachen Jiang, Mohammad Mahdi Khalili, Zhihui Zhu
Abstract: Self-reflection -- the ability of a large language model (LLM) to revisit, evaluate, and revise its own reasoning -- has recently emerged as a powerful behavior enabled by reinforcement learning with verifiable rewards (RLVR). While self-reflection correlates with improved reasoning accuracy, its origin and underlying mechanisms remain poorly understood. In this work, {\it we first show that self-reflection is not exclusive to RLVR fine-tuned models: it already emerges, albeit rarely, in pretrained models}. To probe this latent ability, we introduce Reflection-Inducing Probing, a method that injects reflection-triggering reasoning traces from fine-tuned models into pretrained models. This intervention raises self-reflection frequency of Qwen2.5 from 0.6\% to 18.6\%, revealing a hidden capacity for reflection. Moreover, our analysis of internal representations shows that both pretrained and fine-tuned models maintain hidden states that distinctly separate self-reflective from non-reflective contexts. Leveraging this observation, {\it we then construct a self-reflection vector, a direction in activation space associated with self-reflective reasoning}. By manipulating this vector, we enable bidirectional control over the self-reflective behavior for both pretrained and fine-tuned models. Experiments across multiple reasoning benchmarks show that enhancing these vectors improves reasoning performance by up to 12\%, while suppressing them reduces computational cost, providing a flexible mechanism to navigate the trade-off between reasoning quality and efficiency without requiring additional training. Our findings further our understanding of self-reflection and support a growing body of work showing that understanding model internals can enable precise behavioral control.
Authors: Hantao Yu, Josh Alman
Abstract: The quadratic complexity of self-attention prevents transformers from scaling effectively to long input sequences. On the other hand, modern GPUs and other specialized hardware accelerators are well-optimized for processing small input sequences in transformers during both training and inference. A natural question arises: can we take advantage of the efficiency of small transformers to deal with long input sequences? In this paper, we show that transformers with long input sequences (large transformers) can be efficiently simulated by transformers that can only take short input sequences (small transformers). Specifically, we prove that any transformer with input length $N$ can be efficiently simulated by only $O((N/M)^2)$ transformers with input length $M \ll N$, and that this cannot be improved in the worst case. However, we then prove that in various natural scenarios including average-case inputs, sliding window masking and attention sinks, the optimal number $O(N/M)$ of small transformers suffice.
Authors: Tony Alex, Sara Ahmed, Armin Mustafa, Muhammad Awais, Philip JB Jackson
Abstract: Self-supervised pre-trained audio networks have seen widespread adoption in real-world systems, particularly in multi-modal large language models. These networks are often employed in a frozen state, under the assumption that the SSL pre-training has sufficiently equipped them to handle real-world audio. However, a critical question remains: how well do these models actually perform in real-world conditions, where audio is typically polyphonic and complex, involving multiple overlapping sound sources? Current audio SSL methods are often benchmarked on datasets predominantly featuring monophonic audio, such as environmental sounds, and speech. As a result, the ability of SSL models to generalize to polyphonic audio, a common characteristic in natural scenarios, remains underexplored. This limitation raises concerns about the practical robustness of SSL models in more realistic audio settings. To address this gap, we introduce Self-Supervised Learning from Audio Mixtures (SSLAM), a novel direction in audio SSL research, designed to improve, designed to improve the model's ability to learn from polyphonic data while maintaining strong performance on monophonic data. We thoroughly evaluate SSLAM on standard audio SSL benchmark datasets which are predominantly monophonic and conduct a comprehensive comparative analysis against SOTA methods using a range of high-quality, publicly available polyphonic datasets. SSLAM not only improves model performance on polyphonic audio, but also maintains or exceeds performance on standard audio SSL benchmarks. Notably, it achieves up to a 3.9\% improvement on the AudioSet-2M (AS-2M), reaching a mean average precision (mAP) of 50.2. For polyphonic datasets, SSLAM sets new SOTA in both linear evaluation and fine-tuning regimes with performance improvements of up to 9.1\% (mAP).
Authors: Paul S. Rosenbloom, John E. Laird, Christian Lebiere, Andrea Stocco
Abstract: A beginning is made at mapping four neural theories of consciousness onto the Common Model of Cognition. This highlights how the four jointly depend on recurrent local modules plus a cognitive cycle operating on a global working memory with complex states, and reveals how an existing integrative view of consciousness from a neural perspective aligns with the Com-mon Model.
Authors: Khadija Zanna, Akane Sano
Abstract: Causal discovery (CD) plays a pivotal role in understanding the mechanisms underlying complex systems. While recent algorithms can detect spurious associations and latent confounding, many struggle to recover fairness-relevant pathways in realistic, noisy settings. Large Language Models (LLMs), with their access to broad semantic knowledge, offer a promising complement to statistical CD approaches, particularly in domains where metadata provides meaningful relational cues. Ensuring fairness in machine learning requires understanding how sensitive attributes causally influence outcomes, yet CD methods often introduce spurious or biased pathways. We propose a hybrid LLM-based framework for CD that extends a breadth-first search (BFS) strategy with active learning and dynamic scoring. Variable pairs are prioritized for LLM-based querying using a composite score based on mutual information, partial correlation, and LLM confidence, improving discovery efficiency and robustness. To evaluate fairness sensitivity, we construct a semi-synthetic benchmark from the UCI Adult dataset, embedding a domain-informed causal graph with injected noise, label corruption, and latent confounding. We assess how well CD methods recover both global structure and fairness-critical paths. Our results show that LLM-guided methods, including the proposed method, demonstrate competitive or superior performance in recovering such pathways under noisy conditions. We highlight when dynamic scoring and active querying are most beneficial and discuss implications for bias auditing in real-world datasets.
Authors: Tetiana Gladkykh, Kyrylo Kirykov
Abstract: Text-to-SQL systems enable users to query databases using natural language, democratizing access to data analytics. However, they face challenges in understanding ambiguous phrasing, domain-specific vocabulary, and complex schema relationships. This paper introduces Datrics Text2SQL, a Retrieval-Augmented Generation (RAG)-based framework designed to generate accurate SQL queries by leveraging structured documentation, example-based learning, and domain-specific rules. The system builds a rich Knowledge Base from database documentation and question-query examples, which are stored as vector embeddings and retrieved through semantic similarity. It then uses this context to generate syntactically correct and semantically aligned SQL code. The paper details the architecture, training methodology, and retrieval logic, highlighting how the system bridges the gap between user intent and database structure without requiring SQL expertise.
Authors: Eva Paraschou, Ioannis Arapakis, Sofia Yfantidou, Sebastian Macaluso, Athena Vakali
Abstract: Artificial Intelligence (AI) is rapidly embedded in critical decision-making systems, however their foundational ``black-box'' models require eXplainable AI (XAI) solutions to enhance transparency, which are mostly oriented to experts, making no sense to non-experts. Alarming evidence about AI's unprecedented human values risks brings forward the imperative need for transparent human-centered XAI solutions. In this work, we introduce a domain-, model-, explanation-agnostic, generalizable and reproducible framework that ensures both transparency and human-centered explanations tailored to the needs of both experts and non-experts. The framework leverages Large Language Models (LLMs) and employs in-context learning to convey domain- and explainability-relevant contextual knowledge into LLMs. Through its structured prompt and system setting, our framework encapsulates in one response explanations understandable by non-experts and technical information to experts, all grounded in domain and explainability principles. To demonstrate the effectiveness of our framework, we establish a ground-truth contextual ``thesaurus'' through a rigorous benchmarking with over 40 data, model, and XAI combinations for an explainable clustering analysis of a well-being scenario. Through a comprehensive quality and human-friendliness evaluation of our framework's explanations, we prove high content quality through strong correlations with ground-truth explanations (Spearman rank correlation=0.92) and improved interpretability and human-friendliness to non-experts through a user study (N=56). Our overall evaluation confirms trust in LLMs as HCXAI enablers, as our framework bridges the above Gaps by delivering (i) high-quality technical explanations aligned with foundational XAI methods and (ii) clear, efficient, and interpretable human-centered explanations for non-experts.
Authors: Arno Simons, Michael Zichert, Adrian W\"uthrich
Abstract: This paper explores the use of large language models (LLMs) as research tools in the history, philosophy, and sociology of science (HPSS). LLMs are remarkably effective at processing unstructured text and inferring meaning from context, offering new affordances that challenge long-standing divides between computational and interpretive methods. This raises both opportunities and challenges for HPSS, which emphasizes interpretive methodologies and understands meaning as context-dependent, ambiguous, and historically situated. We argue that HPSS is uniquely positioned not only to benefit from LLMs' capabilities but also to interrogate their epistemic assumptions and infrastructural implications. To this end, we first offer a concise primer on LLM architectures and training paradigms tailored to non-technical readers. We frame LLMs not as neutral tools but as epistemic infrastructures that encode assumptions about meaning, context, and similarity, conditioned by their training data, architecture, and patterns of use. We then examine how computational techniques enhanced by LLMs, such as structuring data, detecting patterns, and modeling dynamic processes, can be applied to support interpretive research in HPSS. Our analysis compares full-context and generative models, outlines strategies for domain and task adaptation (e.g., continued pretraining, fine-tuning, and retrieval-augmented generation), and evaluates their respective strengths and limitations for interpretive inquiry in HPSS. We conclude with four lessons for integrating LLMs into HPSS: (1) model selection involves interpretive trade-offs; (2) LLM literacy is foundational; (3) HPSS must define its own benchmarks and corpora; and (4) LLMs should enhance, not replace, interpretive methods.
Authors: Jennifer Grannen, Siddharth Karamcheti, Blake Wulfe, Dorsa Sadigh
Abstract: Collaborative robots must quickly adapt to their partner's intent and preferences to proactively identify helpful actions. This is especially true in situated settings where human partners can continually teach robots new high-level behaviors, visual concepts, and physical skills (e.g., through demonstration), growing the robot's capabilities as the human-robot pair work together to accomplish diverse tasks. In this work, we argue that robots should be able to infer their partner's goals from early interactions and use this information to proactively plan behaviors ahead of explicit instructions from the user. Building from the strong commonsense priors and steerability of large language models, we introduce ProVox ("Proactive Voice"), a novel framework that enables robots to efficiently personalize and adapt to individual collaborators. We design a meta-prompting protocol that empowers users to communicate their distinct preferences, intent, and expected robot behaviors ahead of starting a physical interaction. ProVox then uses the personalized prompt to condition a proactive language model task planner that anticipates a user's intent from the current interaction context and robot capabilities to suggest helpful actions; in doing so, we alleviate user burden, minimizing the amount of time partners spend explicitly instructing and supervising the robot. We evaluate ProVox through user studies grounded in household manipulation tasks (e.g., assembling lunch bags) that measure the efficiency of the collaboration, as well as features such as perceived helpfulness, ease of use, and reliability. Our analysis suggests that both meta-prompting and proactivity are critical, resulting in 38.7% faster task completion times and 31.9% less user burden relative to non-active baselines. Supplementary material, code, and videos can be found at https://provox-2025.github.io.
Authors: Hui Wei, Dong Yoon Lee, Shubham Rohal, Zhizhang Hu, Shiwei Fang, Shijia Pan
Abstract: Foundation models have gained growing interest in the IoT domain due to their reduced reliance on labeled data and strong generalizability across tasks, which address key limitations of traditional machine learning approaches. However, most existing foundation model based methods are developed for specific IoT tasks, making it difficult to compare approaches across IoT domains and limiting guidance for applying them to new tasks. This survey aims to bridge this gap by providing a comprehensive overview of current methodologies and organizing them around four shared performance objectives by different domains: efficiency, context-awareness, safety, and security & privacy. For each objective, we review representative works, summarize commonly-used techniques and evaluation metrics. This objective-centric organization enables meaningful cross-domain comparisons and offers practical insights for selecting and designing foundation model based solutions for new IoT tasks. We conclude with key directions for future research to guide both practitioners and researchers in advancing the use of foundation models in IoT applications.
Authors: Avinash Baidya, Kamalika Das, Xiang Gao
Abstract: Large Language Model (LLM)-based agents have significantly impacted Task-Oriented Dialog Systems (TODS) but continue to face notable performance challenges, especially in zero-shot scenarios. While prior work has noted this performance gap, the behavioral factors driving the performance gap remain under-explored. This study proposes a comprehensive evaluation framework to quantify the behavior gap between AI agents and human experts, focusing on discrepancies in dialog acts, tool usage, and knowledge utilization. Our findings reveal that this behavior gap is a critical factor negatively impacting the performance of LLM agents. Notably, as task complexity increases, the behavior gap widens (correlation: 0.963), leading to a degradation of agent performance on complex task-oriented dialogs. For the most complex task in our study, even the GPT-4o-based agent exhibits low alignment with human behavior, with low F1 scores for dialog acts (0.464), excessive and often misaligned tool usage with a F1 score of 0.139, and ineffective usage of external knowledge. Reducing such behavior gaps leads to significant performance improvement (24.3% on average). This study highlights the importance of comprehensive behavioral evaluations and improved alignment strategies to enhance the effectiveness of LLM-based TODS in handling complex tasks.
Authors: Yinghao Ma, Siyou Li, Juntao Yu, Emmanouil Benetos, Akira Maezawa
Abstract: Recent advances in audio-text large language models (LLMs) have opened new possibilities for music understanding and generation. However, existing benchmarks are limited in scope, often relying on simplified tasks or multi-choice evaluations that fail to reflect the complexity of real-world music analysis. We reinterpret a broad range of traditional MIR annotations as instruction-following formats and introduce CMI-Bench, a comprehensive music instruction following benchmark designed to evaluate audio-text LLMs on a diverse set of music information retrieval (MIR) tasks. These include genre classification, emotion regression, emotion tagging, instrument classification, pitch estimation, key detection, lyrics transcription, melody extraction, vocal technique recognition, instrument performance technique detection, music tagging, music captioning, and (down)beat tracking: reflecting core challenges in MIR research. Unlike previous benchmarks, CMI-Bench adopts standardized evaluation metrics consistent with previous state-of-the-art MIR models, ensuring direct comparability with supervised approaches. We provide an evaluation toolkit supporting all open-source audio-textual LLMs, including LTU, Qwen-audio, SALMONN, MusiLingo, etc. Experiment results reveal significant performance gaps between LLMs and supervised models, along with their culture, chronological and gender bias, highlighting the potential and limitations of current models in addressing MIR tasks. CMI-Bench establishes a unified foundation for evaluating music instruction following, driving progress in music-aware LLMs.
Authors: Taegyeong Lee, Jeonghwa Yoo, Hyoungseo Cho, Soo Yong Kim, Yunho Maeng
Abstract: The recent advancements in Large Language Models(LLMs) have had a significant impact on a wide range of fields, from general domains to specialized areas. However, these advancements have also significantly increased the potential for malicious users to exploit harmful and jailbreak prompts for malicious attacks. Although there have been many efforts to prevent harmful prompts and jailbreak prompts, protecting LLMs from such malicious attacks remains an important and challenging task. In this paper, we propose QGuard, a simple yet effective safety guard method, that utilizes question prompting to block harmful prompts in a zero-shot manner. Our method can defend LLMs not only from text-based harmful prompts but also from multi-modal harmful prompt attacks. Moreover, by diversifying and modifying guard questions, our approach remains robust against the latest harmful prompts without fine-tuning. Experimental results show that our model performs competitively on both text-only and multi-modal harmful datasets. Additionally, by providing an analysis of question prompting, we enable a white-box analysis of user inputs. We believe our method provides valuable insights for real-world LLM services in mitigating security risks associated with harmful prompts.
Authors: Yue Wan, Xiaowei Jia, Xiang Lorraine Li
Abstract: Chain-of-thought (CoT) prompting has been widely adopted to enhance the reasoning capabilities of large language models (LLMs). However, the effectiveness of CoT reasoning is inconsistent across tasks with different reasoning types. This work presents a novel perspective to understand CoT behavior through the lens of \textit{confirmation bias} in cognitive psychology. Specifically, we examine how model internal beliefs, approximated by direct question-answering probabilities, affect both reasoning generation ($Q \to R$) and reasoning-guided answer prediction ($QR \to A$) in CoT. By decomposing CoT into a two-stage process, we conduct a thorough correlation analysis in model beliefs, rationale attributes, and stage-wise performance. Our results provide strong evidence of confirmation bias in LLMs, such that model beliefs not only skew the reasoning process but also influence how rationales are utilized for answer prediction. Furthermore, the interplay between task vulnerability to confirmation bias and the strength of beliefs also provides explanations for CoT effectiveness across reasoning tasks and models. Overall, this study provides a valuable insight for the needs of better prompting strategies that mitigate confirmation bias to enhance reasoning performance. Code is available at \textit{https://github.com/yuewan2/biasedcot}.
Authors: Xiaotian Zhang, Yuan Wang, Zhaopeng Feng, Ruizhe Chen, Zhijie Zhou, Yan Zhang, Hongxia Xu, Jian Wu, Zuozhu Liu
Abstract: Medical Question-Answering (QA) encompasses a broad spectrum of tasks, including multiple choice questions (MCQ), open-ended text generation, and complex computational reasoning. Despite this variety, a unified framework for delivering high-quality medical QA has yet to emerge. Although recent progress in reasoning-augmented large language models (LLMs) has shown promise, their ability to achieve comprehensive medical understanding is still largely unexplored. In this paper, we present Med-U1, a unified framework for robust reasoning across medical QA tasks with diverse output formats, ranging from MCQs to complex generation and computation tasks. Med-U1 employs pure large-scale reinforcement learning with mixed rule-based binary reward functions, incorporating a length penalty to manage output verbosity. With multi-objective reward optimization, Med-U1 directs LLMs to produce concise and verifiable reasoning chains. Empirical results reveal that Med-U1 significantly improves performance across multiple challenging Med-QA benchmarks, surpassing even larger specialized and proprietary models. Furthermore, Med-U1 demonstrates robust generalization to out-of-distribution (OOD) tasks. Extensive analysis presents insights into training strategies, reasoning chain length control, and reward design for medical LLMs. The code will be released.
Authors: Weipeng Jiang, Xiaoyu Zhang, Xiaofei Xie, Jiongchi Yu, Yuhan Zhi, Shiqing Ma, Chao Shen
Abstract: Large Language Model (LLM) libraries have emerged as the foundational infrastructure powering today's AI revolution, serving as the backbone for LLM deployment, inference optimization, fine-tuning, and production serving across diverse applications. Despite their critical role in the LLM ecosystem, these libraries face frequent quality issues and bugs that threaten the reliability of AI systems built upon them. To address this knowledge gap, we present the first comprehensive empirical investigation into bug characteristics and testing practices in modern LLM libraries. We examine 313 bug-fixing commits extracted across two widely-adopted LLM libraries: HuggingFace Transformers and vLLM.Through rigorous manual analysis, we establish comprehensive taxonomies categorizing bug symptoms into 5 types and root causes into 14 distinct categories.Our primary discovery shows that API misuse has emerged as the predominant root cause (32.17%-48.19%), representing a notable transition from algorithm-focused defects in conventional deep learning frameworks toward interface-oriented problems. Additionally, we examine 7,748 test functions to identify 7 distinct test oracle categories employed in current testing approaches, with predefined expected outputs (such as specific tensors and text strings) being the most common strategy. Our assessment of existing testing effectiveness demonstrates that the majority of bugs escape detection due to inadequate test cases (41.73%), lack of test drivers (32.37%), and weak test oracles (25.90%). Drawing from these findings, we offer some recommendations for enhancing LLM library quality assurance.
Authors: Jie Zhang, Qinghua Zhao, Lei Li, Chi-ho Lin
Abstract: Large language models have demonstrated a remarkable ability for verbatim memorization. While numerous works have explored factors influencing model memorization, the dynamic evolution memorization patterns remains underexplored. This paper presents a detailed analysis of memorization in the Pythia model family across varying scales and training steps under prefix perturbations. Using granular metrics, we examine how model architecture, data characteristics, and perturbations influence these patterns. Our findings reveal that: (1) as model scale increases, memorization expands incrementally while efficiency decreases rapidly; (2) as model scale increases, the rate of new memorization acquisition decreases while old memorization forgetting increases; (3) data characteristics (token frequency, repetition count, and uncertainty) differentially affect memorized versus non-memorized samples; and (4) prefix perturbations reduce memorization and increase generation uncertainty proportionally to perturbation strength, with low-redundancy samples showing higher vulnerability and larger models offering no additional robustness. These findings advance our understanding of memorization mechanisms, with direct implications for training optimization, privacy safeguards, and architectural improvements.
Authors: Johnny Peng, Thanh Tung Khuat, Katarzyna Musial, Bogdan Gabrys
Abstract: Data is crucial for machine learning (ML) applications, yet acquiring large datasets can be costly and time-consuming, especially in complex, resource-intensive fields like biopharmaceuticals. A key process in this industry is upstream bioprocessing, where living cells are cultivated and optimised to produce therapeutic proteins and biologics. The intricate nature of these processes, combined with high resource demands, often limits data collection, resulting in smaller datasets. This comprehensive review explores ML methods designed to address the challenges posed by small data and classifies them into a taxonomy to guide practical applications. Furthermore, each method in the taxonomy was thoroughly analysed, with a detailed discussion of its core concepts and an evaluation of its effectiveness in tackling small data challenges, as demonstrated by application results in the upstream bioprocessing and other related domains. By analysing how these methods tackle small data challenges from different perspectives, this review provides actionable insights, identifies current research gaps, and offers guidance for leveraging ML in data-constrained environments.
Authors: Yongmin Kwon, Namwoo Kang
Abstract: Generative models have attracted considerable attention for their ability to produce novel shapes. However, their application in mechanical design remains constrained due to the limited size and variability of available datasets. This study proposes a deep learning-based optimization framework specifically tailored for shape optimization with limited datasets, leveraging positional encoding and a Lipschitz regularization term to robustly learn geometric characteristics and maintain a meaningful latent space. Through extensive experiments, the proposed approach demonstrates robustness, generalizability and effectiveness in addressing typical limitations of conventional optimization frameworks. The validity of the methodology is confirmed through multi-objective shape optimization experiments conducted on diverse three-dimensional datasets, including wheels and cars, highlighting the model's versatility in producing practical and high-quality design outcomes even under data-constrained conditions.
Authors: Hitomi Yanaka, Xinqi He, Jie Lu, Namgi Han, Sunjin Oh, Ryoma Kumon, Yuma Matsuoka, Katsuhiko Watabe, Yuko Itatsu
Abstract: An growing number of studies have examined the social bias of rapidly developed large language models (LLMs). Although most of these studies have focused on bias occurring in a single social attribute, research in social science has shown that social bias often occurs in the form of intersectionality -- the constitutive and contextualized perspective on bias aroused by social attributes. In this study, we construct the Japanese benchmark inter-JBBQ, designed to evaluate the intersectional bias in LLMs on the question-answering setting. Using inter-JBBQ to analyze GPT-4o and Swallow, we find that biased output varies according to its contexts even with the equal combination of social attributes.
Authors: Dekun Wu, Frederik Brudy, Bang Liu, Yi Wang
Abstract: Virtual environments are essential to AI agent research. Existing environments for LLM agent research typically focus on either physical task solving or social simulation, with the former oversimplifying agent individuality and social dynamics, and the latter lacking physical grounding of social behaviors. We introduce IndoorWorld, a heterogeneous multi-agent environment that tightly integrates physical and social dynamics. By introducing novel challenges for LLM-driven agents in orchestrating social dynamics to influence physical environments and anchoring social interactions within world states, IndoorWorld opens up possibilities of LLM-based building occupant simulation for architectural design. We demonstrate the potential with a series of experiments within an office setting to examine the impact of multi-agent collaboration, resource competition, and spatial layout on agent behavior.
Authors: Chuntao Ding, Jianhang Xie, Junna Zhang, Salman Raza, Shangguang Wang, Jiannong Cao
Abstract: It has become mainstream to deploy Convolutional Neural Network (CNN) models on ubiquitous Internet of Things (IoT) devices with the help of the cloud to provide users with a variety of high-quality services. Most existing methods have two limitations: (i) low robustness in handling corrupted image data collected by IoT devices; and (ii) high consumption of computational and transmission resources. To this end, we propose the Grouped NonLinear transformation generation method (GroupNL), which generates diversified feature maps by utilizing data-agnostic Nonlinear Transformation Functions (NLFs) to improve the robustness of the CNN model. Specifically, partial convolution filters are designated as seed filters in a convolutional layer, and a small set of feature maps, i.e., seed feature maps, are first generated based on vanilla convolution operation. Then, we split seed feature maps into several groups, each with a set of different NLFs, to generate corresponding diverse feature maps with in-place nonlinear processing. Moreover, GroupNL effectively reduces the parameter transmission between multiple nodes during model training by setting the hyperparameters of NLFs to random initialization and not updating them during model training, and reduces the computing resources by using NLFs to generate feature maps instead of most feature maps generated based on sliding windows. Experimental results on CIFAR-10, GTSRB, CIFAR-10-C, Icons50, and ImageNet-1K datasets in NVIDIA RTX GPU platforms show that the proposed GroupNL outperforms other state-of-the-art methods in model robust and training acceleration. Specifically, on the Icons-50 dataset, the accuracy of GroupNL-ResNet-18 achieves approximately 2.86% higher than the vanilla ResNet-18. GroupNL improves training speed by about 53% compared to vanilla CNN when trained on a cluster of 8 NVIDIA RTX 4090 GPUs on the ImageNet-1K dataset.
Authors: Ruiyan Zhu, Xi Cheng, Ke Liu, Brian Zhu, Daniel Jin, Neeraj Parihar, Zhoutian Xu, Oliver Gao
Abstract: We present SheetMind, a modular multi-agent framework powered by large language models (LLMs) for spreadsheet automation via natural language instructions. The system comprises three specialized agents: a Manager Agent that decomposes complex user instructions into subtasks; an Action Agent that translates these into structured commands using a Backus Naur Form (BNF) grammar; and a Reflection Agent that validates alignment between generated actions and the user's original intent. Integrated into Google Sheets via a Workspace extension, SheetMind supports real-time interaction without requiring scripting or formula knowledge. Experiments on benchmark datasets demonstrate an 80 percent success rate on single step tasks and approximately 70 percent on multi step instructions, outperforming ablated and baseline variants. Our results highlight the effectiveness of multi agent decomposition and grammar based execution for bridging natural language and spreadsheet functionalities.
Authors: Arjun R. Akula, Kazuma Hashimoto, Krishna Srinivasan, Aditi Chaudhary, Karthik Raman, Michael Bendersky
Abstract: The emergence of long-context large language models (LLMs) has enabled the use of hundreds, or even thousands, of demonstrations for in-context learning (ICL) - a previously impractical regime. This paper investigates whether traditional ICL selection strategies, which balance the similarity of ICL examples to the test input (using a text retriever) with diversity within the ICL set, remain effective when utilizing a large number of demonstrations. Our experiments demonstrate that, while longer contexts can accommodate more examples, simply increasing the number of demonstrations does not guarantee improved performance. Smart ICL selection remains crucial, even with thousands of demonstrations. To further enhance ICL in this setting, we introduce Refract ICL, a novel ICL selection algorithm specifically designed to focus LLM attention on challenging examples by strategically repeating them within the context and incorporating zero-shot predictions as error signals. Our results show that Refract ICL significantly improves the performance of extremely long-context models such as Gemini 1.5 Pro, particularly on tasks with a smaller number of output classes.
Authors: Peiran Qiu, Siyi Zhou, Emilio Ferrara
Abstract: This study examines information suppression mechanisms in DeepSeek, an open-source large language model (LLM) developed in China. We propose an auditing framework and use it to analyze the model's responses to 646 politically sensitive prompts by comparing its final output with intermediate chain-of-thought (CoT) reasoning. Our audit unveils evidence of semantic-level information suppression in DeepSeek: sensitive content often appears within the model's internal reasoning but is omitted or rephrased in the final output. Specifically, DeepSeek suppresses references to transparency, government accountability, and civic mobilization, while occasionally amplifying language aligned with state propaganda. This study underscores the need for systematic auditing of alignment, content moderation, information suppression, and censorship practices implemented into widely-adopted AI models, to ensure transparency, accountability, and equitable access to unbiased information obtained by means of these systems.
Authors: Jiancong Xiao, Zhekun Shi, Kaizhao Liu, Qi Long, Weijie J. Su
Abstract: Despite its empirical success, Reinforcement Learning from Human Feedback (RLHF) has been shown to violate almost all the fundamental axioms in social choice theory -- such as majority consistency, pairwise majority consistency, and Condorcet consistency. This raises a foundational question: why does RLHF perform so well in practice if it fails these seemingly essential properties? In this paper, we resolve this paradox by showing that under mild and empirically plausible assumptions on the preference profile, RLHF does satisfy pairwise majority and Condorcet consistency. These assumptions are frequently satisfied in real-world alignment tasks, offering a theoretical explanation for RLHF's strong practical performance. Furthermore, we show that a slight modification to the reward modeling objective can ensure pairwise majority or Condorcet consistency even under general preference profiles, thereby improving the alignment process. Finally, we go beyond classical axioms in economic and social choice theory and introduce new alignment criteria -- preference matching, preference equivalence, and group preference matching -- that better reflect the goal of learning distributions over responses. We show that while RLHF satisfies the first two properties, it fails to satisfy the third. We conclude by discussing how future alignment methods may be designed to satisfy all three.
Authors: Kaiyuan Liu, Chen Shen, Zhanwei Zhang, Junjie Liu, Xiaosong Yuan, Jieping ye
Abstract: While recent advances in large reasoning models have demonstrated remarkable performance, efficient reasoning remains critical due to the rapid growth of output length. Existing optimization approaches highlights a tendency toward "overthinking", yet lack fine-grained analysis. In this work, we focus on Self-Affirmation Reflections: redundant reflective steps that affirm prior content and often occurs after the already correct reasoning steps. Observations of both original and optimized reasoning models reveal pervasive self-affirmation reflections. Notably, these reflections sometimes lead to longer outputs in optimized models than their original counterparts. Through detailed analysis, we uncover an intriguing pattern: compared to other reflections, the leading words (i.e., the first word of sentences) in self-affirmation reflections exhibit a distinct probability bias. Motivated by this insight, we can locate self-affirmation reflections and conduct a train-free experiment demonstrating that suppressing self-affirmation reflections reduces output length without degrading accuracy across multiple models (R1-Distill-Models, QwQ-32B, and Qwen3-32B). Furthermore, we also improve current train-based method by explicitly suppressing such reflections. In our experiments, we achieve length compression of 18.7\% in train-free settings and 50.2\% in train-based settings for R1-Distill-Qwen-1.5B. Moreover, our improvements are simple yet practical and can be directly applied to existing inference frameworks, such as vLLM. We believe that our findings will provide community insights for achieving more precise length compression and step-level efficient reasoning.
Authors: Xingyue Huang, Mikhail Galkin, Michael M. Bronstein, \.Ismail \.Ilkan Ceylan
Abstract: Inductive link prediction with knowledge hypergraphs is the task of predicting missing hyperedges involving completely novel entities (i.e., nodes unseen during training). Existing methods for inductive link prediction with knowledge hypergraphs assume a fixed relational vocabulary and, as a result, cannot generalize to knowledge hypergraphs with novel relation types (i.e., relations unseen during training). Inspired by knowledge graph foundation models, we propose HYPER as a foundation model for link prediction, which can generalize to any knowledge hypergraph, including novel entities and novel relations. Importantly, HYPER can learn and transfer across different relation types of varying arities, by encoding the entities of each hyperedge along with their respective positions in the hyperedge. To evaluate HYPER, we construct 16 new inductive datasets from existing knowledge hypergraphs, covering a diverse range of relation types of varying arities. Empirically, HYPER consistently outperforms all existing methods in both node-only and node-and-relation inductive settings, showing strong generalization to unseen, higher-arity relational structures.
Authors: Wenbo Li, Shiyi Wang, Yiteng Chen, Huiping Zhuang, Qingyao Wu
Abstract: Vision-Language Models (VLMs) encode knowledge and reasoning capabilities for robotic manipulation within high-dimensional representation spaces. However, current approaches often project them into compressed intermediate representations, discarding important task-specific information such as fine-grained spatial or semantic details. To address this, we propose AntiGrounding, a new framework that reverses the instruction grounding process. It lifts candidate actions directly into the VLM representation space, renders trajectories from multiple views, and uses structured visual question answering for instruction-based decision making. This enables zero-shot synthesis of optimal closed-loop robot trajectories for new tasks. We also propose an offline policy refinement module that leverages past experience to enhance long-term performance. Experiments in both simulation and real-world environments show that our method outperforms baselines across diverse robotic manipulation tasks.
Authors: Stan Mu\~noz Guti\'errez, Franz Wotawa
Abstract: This paper introduces Spectral Fault Receptive Fields (SFRFs), a biologically inspired technique for degradation state assessment in bearing fault diagnosis and remaining useful life (RUL) estimation. Drawing on the center-surround organization of retinal ganglion cell receptive fields, we propose a frequency-domain feature extraction algorithm that enhances the detection of fault signatures in vibration signals. SFRFs are designed as antagonistic spectral filters centered on characteristic fault frequencies, with inhibitory surrounds that enable robust characterization of incipient faults under variable operating conditions. A multi-objective evolutionary optimization strategy based on NSGA-II algorithm is employed to tune the receptive field parameters by simultaneously minimizing RUL prediction error, maximizing feature monotonicity, and promoting smooth degradation trajectories. The method is demonstrated on the XJTU-SY bearing run-to-failure dataset, confirming its suitability for constructing condition indicators in health monitoring applications. Key contributions include: (i) the introduction of SFRFs, inspired by the biology of vision in the primate retina; (ii) an evolutionary optimization framework guided by condition monitoring and prognosis criteria; and (iii) experimental evidence supporting the detection of early-stage faults and their precursors. Furthermore, we confirm that our diagnosis-informed spectral representation achieves accurate RUL prediction using a bagging regressor. The results highlight the interpretability and principled design of SFRFs, bridging signal processing, biological sensing principles, and data-driven prognostics in rotating machinery.
Authors: Barra White, Krishnendu Guha
Abstract: Explainable ML algorithms are designed to provide transparency and insight into their decision-making process. Explaining how ML models come to their prediction is critical in fields such as healthcare and finance, as it provides insight into how models can help detect bias in predictions and help comply with GDPR compliance in these fields. QML leverages quantum phenomena such as entanglement and superposition, offering the potential for computational speedup and greater insights compared to classical ML. However, QML models also inherit the black-box nature of their classical counterparts, requiring the development of explainability techniques to be applied to these QML models to help understand why and how a particular output was generated. This paper will explore the idea of creating a modular, explainable QML framework that splits QML algorithms into their core components, such as feature maps, variational circuits (ansatz), optimizers, kernels, and quantum-classical loops. Each component will be analyzed using explainability techniques, such as ALE and SHAP, which have been adapted to analyse the different components of these QML algorithms. By combining insights from these parts, the paper aims to infer explainability to the overall QML model.
Authors: Zichuan Fu, Xian Wu, Yejing Wang, Wanyu Wang, Shanshan Ye, Hongzhi Yin, Yi Chang, Yefeng Zheng, Xiangyu Zhao
Abstract: Large Language Models (LLMs) have demonstrated exceptional capabilities across diverse natural language processing (NLP) tasks. The release of open-source LLMs like LLaMA and Qwen has triggered the development of numerous fine-tuned models tailored for various tasks and languages. In this paper, we explore an important question: is it possible to combine these specialized models to create a unified model with multi-task capabilities. We introduces Hierarchical Iterative Merging (Hi-Merging), a training-free method for unifying different specialized LLMs into a single model. Specifically, Hi-Merging employs model-wise and layer-wise pruning and scaling, guided by contribution analysis, to mitigate parameter conflicts. Extensive experiments on multiple-choice and question-answering tasks in both Chinese and English validate Hi-Merging's ability for multi-task learning. The results demonstrate that Hi-Merging consistently outperforms existing merging techniques and surpasses the performance of models fine-tuned on combined datasets in most scenarios. Code is available at: https://github.com/Applied-Machine-Learning-Lab/Hi-Merging.
URLs: https://github.com/Applied-Machine-Learning-Lab/Hi-Merging.
Authors: Jiawei Chen, Zhengwei Fang, Xiao Yang, Chao Yu, Zhaoxia Yin, Hang Su
Abstract: Ensuring the safety and alignment of Large Language Models is a significant challenge with their growing integration into critical applications and societal functions. While prior research has primarily focused on jailbreak attacks, less attention has been given to non-adversarial failures that subtly emerge during benign interactions. We introduce secondary risks a novel class of failure modes marked by harmful or misleading behaviors during benign prompts. Unlike adversarial attacks, these risks stem from imperfect generalization and often evade standard safety mechanisms. To enable systematic evaluation, we introduce two risk primitives verbose response and speculative advice that capture the core failure patterns. Building on these definitions, we propose SecLens, a black-box, multi-objective search framework that efficiently elicits secondary risk behaviors by optimizing task relevance, risk activation, and linguistic plausibility. To support reproducible evaluation, we release SecRiskBench, a benchmark dataset of 650 prompts covering eight diverse real-world risk categories. Experimental results from extensive evaluations on 16 popular models demonstrate that secondary risks are widespread, transferable across models, and modality independent, emphasizing the urgent need for enhanced safety mechanisms to address benign yet harmful LLM behaviors in real-world deployments.
Authors: Andrej Kastrin, Bojan Cestnik, Nada Lavra\v{c}
Abstract: The explosive growth of scientific publications has created an urgent need for automated methods that facilitate knowledge synthesis and hypothesis generation. Literature-based discovery (LBD) addresses this challenge by uncovering previously unknown associations between disparate domains. This article surveys recent methodological advances in LBD, focusing on developments from 2000 to the present. We review progress in three key areas: knowledge graph construction, deep learning approaches, and the integration of pre-trained and large language models (LLMs). While LBD has made notable progress, several fundamental challenges remain unresolved, particularly concerning scalability, reliance on structured data, and the need for extensive manual curation. By examining ongoing advances and outlining promising future directions, this survey underscores the transformative role of LLMs in enhancing LBD and aims to support researchers and practitioners in harnessing these technologies to accelerate scientific innovation.
Authors: Chong Li, Yingzhuo Deng, Jiajun Zhang, Chengqing Zong
Abstract: The curse of multilinguality phenomenon is a fundamental problem of multilingual Large Language Models (LLMs), where the competition between massive languages results in inferior performance. It mainly comes from limited capacity and negative transfer between dissimilar languages. To address this issue, we propose a method to dynamically group and scale up the parameters of multilingual LLM while boosting positive transfer among similar languages. Specifically, the model is first tuned on monolingual corpus to determine the parameter deviation in each layer and quantify the similarity between languages. Layers with more deviations are extended to mixture-of-experts layers to reduce competition between languages, where one expert module serves one group of similar languages. Experimental results on 18 to 128 languages show that our method reduces the negative transfer between languages and significantly boosts multilingual performance with fewer parameters. Such language group specialization on experts benefits the new language adaptation and reduces the inference on the previous multilingual knowledge learned.
Authors: Zhiyuan Su, Sunhao Dai, Xiao Zhang
Abstract: Clustering of Bandits (CB) methods enhance sequential decision-making by grouping bandits into clusters based on similarity and incorporating cluster-level contextual information, demonstrating effectiveness and adaptability in applications like personalized streaming recommendations. However, when extending CB algorithms to their neural version (commonly referred to as Clustering of Neural Bandits, or CNB), they suffer from loss of plasticity, where neural network parameters become rigid and less adaptable over time, limiting their ability to adapt to non-stationary environments (e.g., dynamic user preferences in recommendation). To address this challenge, we propose Selective Reinitialization (SeRe), a novel bandit learning framework that dynamically preserves the adaptability of CNB algorithms in evolving environments. SeRe leverages a contribution utility metric to identify and selectively reset underutilized units, mitigating loss of plasticity while maintaining stable knowledge retention. Furthermore, when combining SeRe with CNB algorithms, the adaptive change detection mechanism adjusts the reinitialization frequency according to the degree of non-stationarity, ensuring effective adaptation without unnecessary resets. Theoretically, we prove that SeRe enables sublinear cumulative regret in piecewise-stationary environments, outperforming traditional CNB approaches in long-term performances. Extensive experiments on six real-world recommendation datasets demonstrate that SeRe-enhanced CNB algorithms can effectively mitigate the loss of plasticity with lower regrets, improving adaptability and robustness in dynamic settings.
Authors: Haotian Zhang, Liu Liu, Baosheng Yu, Jiayan Qiu, Yanwei Ren, Xianglong Liu
Abstract: The advent of parameter-efficient fine-tuning methods has significantly reduced the computational burden of adapting large-scale pretrained models to diverse downstream tasks. However, existing approaches often struggle to achieve robust performance under domain shifts while maintaining computational efficiency. To address this challenge, we propose Low-rAnk Regulated Gradient Projection (LARGO) algorithm that integrates dynamic constraints into low-rank adaptation methods. Specifically, LARGO incorporates parallel trainable gradient projections to dynamically regulate layer-wise updates, retaining the Out-Of-Distribution robustness of pretrained model while preserving inter-layer independence. Additionally, it ensures computational efficiency by mitigating the influence of gradient dependencies across layers during weight updates. Besides, through leveraging singular value decomposition of pretrained weights for structured initialization, we incorporate an SVD-based initialization strategy that minimizing deviation from pretrained knowledge. Through extensive experiments on diverse benchmarks, LARGO achieves state-of-the-art performance across in-domain and out-of-distribution scenarios, demonstrating improved robustness under domain shifts with significantly lower computational overhead compared to existing PEFT methods. The source code will be released soon.
Authors: Asghar Ghorbani, Hanieh Fattahi
Abstract: Limited infrastructure, scarce educational resources, and unreliable internet access often hinder physics and photonics education in underdeveloped regions. These barriers create deep inequities in Science, Technology, Engineering, and Mathematics (STEM) education. This article explores how Small Language Models (SLMs)-compact, AI-powered tools that can run offline on low-power devices, offering a scalable solution. By acting as virtual tutors, enabling native-language instruction, and supporting interactive learning, SLMs can help address the shortage of trained educators and laboratory access. By narrowing the digital divide through targeted investment in AI technologies, SLMs present a scalable and inclusive solution to advance STEM education and foster scientific empowerment in marginalized communities.
Authors: Tushar Talukder Showrav, Soyabul Islam Lincoln, Md. Kamrul Hasan
Abstract: Background: Deep learning has significantly advanced ECG arrhythmia classification, enabling high accuracy in detecting various cardiac conditions. The use of single-lead ECG systems is crucial for portable devices, as they offer convenience and accessibility for continuous monitoring in diverse settings. However, the interpretability and reliability of deep learning models in clinical applications poses challenges due to their black-box nature. Methods: To address these challenges, we propose EXGnet, a single-lead, trustworthy ECG arrhythmia classification network that integrates multiresolution feature extraction with Explainable Artificial Intelligence (XAI) guidance and train only quantitative features. Results: Trained on two public datasets, including Chapman and Ningbo, EXGnet demonstrates superior performance through key metrics such as Accuracy, F1-score, Sensitivity, and Specificity. The proposed method achieved average five fold accuracy of 98.762%, and 96.932% and average F1-score of 97.910%, and 95.527% on the Chapman and Ningbo datasets, respectively. Conclusions: By employing XAI techniques, specifically Grad-CAM, the model provides visual insights into the relevant ECG segments it analyzes, thereby enhancing clinician trust in its predictions. While quantitative features further improve classification performance, they are not required during testing, making the model suitable for real-world applications. Overall, EXGnet not only achieves better classification accuracy but also addresses the critical need for interpretability in deep learning, facilitating broader adoption in portable ECG monitoring.
Authors: Vivek Chavan, Arsen Cenaj, Shuyuan Shen, Ariane Bar, Srishti Binwani, Tommaso Del Becaro, Marius Funk, Lynn Greschner, Roberto Hung, Stina Klein, Romina Kleiner, Stefanie Krause, Sylwia Olbrych, Vishvapalsinhji Parmar, Jaleh Sarafraz, Daria Soroko, Daksitha Withanage Don, Chang Zhou, Hoang Thuy Duong Vu, Parastoo Semnani, Daniel Weinhardt, Elisabeth Andre, J\"org Kr\"uger, Xavier Fresquet
Abstract: This paper explores the growing presence of emotionally responsive artificial intelligence through a critical and interdisciplinary lens. Bringing together the voices of early-career researchers from multiple fields, it explores how AI systems that simulate or interpret human emotions are reshaping our interactions in areas such as education, healthcare, mental health, caregiving, and digital life. The analysis is structured around four central themes: the ethical implications of emotional AI, the cultural dynamics of human-machine interaction, the risks and opportunities for vulnerable populations, and the emerging regulatory, design, and technical considerations. The authors highlight the potential of affective AI to support mental well-being, enhance learning, and reduce loneliness, as well as the risks of emotional manipulation, over-reliance, misrepresentation, and cultural bias. Key challenges include simulating empathy without genuine understanding, encoding dominant sociocultural norms into AI systems, and insufficient safeguards for individuals in sensitive or high-risk contexts. Special attention is given to children, elderly users, and individuals with mental health challenges, who may interact with AI in emotionally significant ways. However, there remains a lack of cognitive or legal protections which are necessary to navigate such engagements safely. The report concludes with ten recommendations, including the need for transparency, certification frameworks, region-specific fine-tuning, human oversight, and longitudinal research. A curated supplementary section provides practical tools, models, and datasets to support further work in this domain.
Authors: Federico Simonetta
Abstract: This paper presents the first comprehensive systematic review of literature on style-based composer identification and authorship attribution in symbolic music scores. Addressing the critical need for improved reliability and reproducibility in this field, the review rigorously analyzes 58 peer-reviewed papers published across various historical periods, with the search adapted to evolving terminology. The analysis critically assesses prevailing repertoires, computational approaches, and evaluation methodologies, highlighting significant challenges. It reveals that a substantial portion of existing research suffers from inadequate validation protocols and an over-reliance on simple accuracy metrics for often imbalanced datasets, which can undermine the credibility of attribution claims. The crucial role of robust metrics like Balanced Accuracy and rigorous cross-validation in ensuring trustworthy results is emphasized. The survey also details diverse feature representations and the evolution of machine learning models employed. Notable real-world authorship attribution cases, such as those involving works attributed to Bach, Josquin Desprez, and Lennon-McCartney, are specifically discussed, illustrating the opportunities and pitfalls of applying computational techniques to resolve disputed musical provenance. Based on these insights, a set of actionable guidelines for future research are proposed. These recommendations are designed to significantly enhance the reliability, reproducibility, and musicological validity of composer identification and authorship attribution studies, fostering more robust and interpretable computational stylistic analysis.
Authors: Caixu Xu, Junming Wei, Huizhen Chen, Pengchen Liang, Bocheng Liang, Ying Tan, Xintong Wei
Abstract: Recently, Mamba-based methods have become popular in medical image segmentation due to their lightweight design and long-range dependency modeling capabilities. However, current segmentation methods frequently encounter challenges in fetal ultrasound images, such as enclosed anatomical structures, blurred boundaries, and small anatomical structures. To address the need for balancing local feature extraction and global context modeling, we propose MS-UMamba, a novel hybrid convolutional-mamba model for fetal ultrasound image segmentation. Specifically, we design a visual state space block integrated with a CNN branch (SS-MCAT-SSM), which leverages Mamba's global modeling strengths and convolutional layers' local representation advantages to enhance feature learning. In addition, we also propose an efficient multi-scale feature fusion module that integrates spatial attention mechanisms, which Integrating feature information from different layers enhances the feature representation ability of the model. Finally, we conduct extensive experiments on a non-public dataset, experimental results demonstrate that MS-UMamba model has excellent performance in segmentation performance.
Authors: Jinming Luo, Hailin Wang
Abstract: Relation Extraction (RE) aims to extract semantic relationships in texts from given entity pairs, and has achieved significant improvements. However, in different domains, the RE task can be influenced by various factors. For example, in the financial domain, sentiment can affect RE results, yet this factor has been overlooked by modern RE models. To address this gap, this paper proposes a Sentiment-aware-SDP-Enhanced-Module (SSDP-SEM), a multi-task learning approach for enhancing financial RE. Specifically, SSDP-SEM integrates the RE models with a pluggable auxiliary sentiment perception (ASP) task, enabling the RE models to concurrently navigate their attention weights with the text's sentiment. We first generate detailed sentiment tokens through a sentiment model and insert these tokens into an instance. Then, the ASP task focuses on capturing nuanced sentiment information through predicting the sentiment token positions, combining both sentiment insights and the Shortest Dependency Path (SDP) of syntactic information. Moreover, this work employs a sentiment attention information bottleneck regularization method to regulate the reasoning process. Our experiment integrates this auxiliary task with several prevalent frameworks, and the results demonstrate that most previous models benefit from the auxiliary task, thereby achieving better results. These findings highlight the importance of effectively leveraging sentiment in the financial RE task.
Authors: Chengqing Yu, Fei Wang, Chuanguang Yang, Zezhi Shao, Tao Sun, Tangwen Qian, Wei Wei, Zhulin An, Yongjun Xu
Abstract: Multivariate Time Series Forecasting (MTSF) involves predicting future values of multiple interrelated time series. Recently, deep learning-based MTSF models have gained significant attention for their promising ability to mine semantics (global and local information) within MTS data. However, these models are pervasively susceptible to missing values caused by malfunctioning data collectors. These missing values not only disrupt the semantics of MTS, but their distribution also changes over time. Nevertheless, existing models lack robustness to such issues, leading to suboptimal forecasting performance. To this end, in this paper, we propose Multi-View Representation Learning (Merlin), which can help existing models achieve semantic alignment between incomplete observations with different missing rates and complete observations in MTS. Specifically, Merlin consists of two key modules: offline knowledge distillation and multi-view contrastive learning. The former utilizes a teacher model to guide a student model in mining semantics from incomplete observations, similar to those obtainable from complete observations. The latter improves the student model's robustness by learning from positive/negative data pairs constructed from incomplete observations with different missing rates, ensuring semantic alignment across different missing rates. Therefore, Merlin is capable of effectively enhancing the robustness of existing models against unfixed missing rates while preserving forecasting accuracy. Experiments on four real-world datasets demonstrate the superiority of Merlin.
Authors: Suyeon Kim, SeongKu Kang, Dongwoo Kim, Jungseul Ok, Hwanjo Yu
Abstract: Graph Neural Networks (GNNs) have achieved state-of-the-art performance in node classification tasks but struggle with label noise in real-world data. Existing studies on graph learning with label noise commonly rely on class-dependent label noise, overlooking the complexities of instance-dependent noise and falling short of capturing real-world corruption patterns. We introduce BeGIN (Benchmarking for Graphs with Instance-dependent Noise), a new benchmark that provides realistic graph datasets with various noise types and comprehensively evaluates noise-handling strategies across GNN architectures, noisy label detection, and noise-robust learning. To simulate instance-dependent corruptions, BeGIN introduces algorithmic methods and LLM-based simulations. Our experiments reveal the challenges of instance-dependent noise, particularly LLM-based corruption, and underscore the importance of node-specific parameterization to enhance GNN robustness. By comprehensively evaluating noise-handling strategies, BeGIN provides insights into their effectiveness, efficiency, and key performance factors. We expect that BeGIN will serve as a valuable resource for advancing research on label noise in graphs and fostering the development of robust GNN training methods. The code is available at https://github.com/kimsu55/BeGIN.
Authors: K. J. Kevin Feng, David W. McDonald, Amy X. Zhang
Abstract: Autonomy is a double-edged sword for AI agents, simultaneously unlocking transformative possibilities and serious risks. How can agent developers calibrate the appropriate levels of autonomy at which their agents should operate? We argue that an agent's level of autonomy can be treated as a deliberate design decision, separate from its capability and operational environment. In this work, we define five levels of escalating agent autonomy, characterized by the roles a user can take when interacting with an agent: operator, collaborator, consultant, approver, and observer. Within each level, we describe the ways by which a user can exert control over the agent and open questions for how to design the nature of user-agent interaction. We then highlight a potential application of our framework towards AI autonomy certificates to govern agent behavior in single- and multi-agent systems. We conclude by proposing early ideas for evaluating agents' autonomy. Our work aims to contribute meaningful, practical steps towards responsibly deployed and useful AI agents in the real world.
Authors: Wenyun Li, Wenjie Huang, Zejian Deng, Chen Sun
Abstract: Accurate driving behavior modeling is fundamental to safe and efficient trajectory prediction, yet remains challenging in complex traffic scenarios. This paper presents a novel Inverse Reinforcement Learning (IRL) framework that captures human-like decision-making by inferring diverse reward functions, enabling robust cross-scenario adaptability. The learned reward function is utilized to maximize the likelihood of output by the encoder-decoder architecture that combines Mamba blocks for efficient long-sequence dependency modeling with graph attention networks to encode spatial interactions among traffic agents. Comprehensive evaluations on urban intersections and roundabouts demonstrate that the proposed method not only outperforms various popular approaches in prediction accuracy but also achieves 2 times higher generalization performance to unseen scenarios compared to other IRL-based method.
Authors: Filip Sondej, Yushi Yang, Miko{\l}aj Kniejski, Marcel Windys
Abstract: Language models can retain dangerous knowledge and skills even after extensive safety fine-tuning, posing both misuse and misalignment risks. Recent studies show that even specialized unlearning methods can be easily reversed. To address this, we systematically evaluate many existing and novel components of unlearning methods and identify ones crucial for irreversible unlearning. We introduce Disruption Masking, a technique in which we only allow updating weights, where the signs of the unlearning gradient and the retaining gradient are the same. This ensures all updates are non-disruptive. Additionally, we identify the need for normalizing the unlearning gradients, and also confirm the usefulness of meta-learning. We combine these insights into MUDMAN (Meta-Unlearning with Disruption Masking and Normalization) and validate its effectiveness at preventing the recovery of dangerous capabilities. MUDMAN outperforms the prior TAR method by 40\%, setting a new state-of-the-art for robust unlearning.
Authors: Yanqiao Zhu
Abstract: This paper presents a comparative analysis of deep learning strategies for detecting hypertensive retinopathy from fundus images, a central task in the HRDC challenge~\cite{qian2025hrdc}. We investigate three distinct approaches: a custom CNN, a suite of pre-trained transformer-based models, and an AutoML solution. Our findings reveal a stark, architecture-dependent response to data augmentation. Augmentation significantly boosts the performance of pure Vision Transformers (ViTs), which we hypothesize is due to their weaker inductive biases, forcing them to learn robust spatial and structural features. Conversely, the same augmentation strategy degrades the performance of hybrid ViT-CNN models, whose stronger, pre-existing biases from the CNN component may be "confused" by the transformations. We show that smaller patch sizes (ViT-B/8) excel on augmented data, enhancing fine-grained detail capture. Furthermore, we demonstrate that a powerful self-supervised model like DINOv2 fails on the original, limited dataset but is "rescued" by augmentation, highlighting the critical need for data diversity to unlock its potential. Preliminary tests with a ViT-Large model show poor performance, underscoring the risk of using overly-capacitive models on specialized, smaller datasets. This work provides critical insights into the interplay between model architecture, data augmentation, and dataset size for medical image classification.
Authors: Julie Bauer, Rishabh Kaushal, Thales Bertaglia, Adriana Iamnitchi
Abstract: Numerous studies have proposed computational methods to detect hate speech online, yet most focus on the English language and emphasize model development. In this study, we evaluate the counterfactual fairness of hate speech detection models in the Dutch language, specifically examining the performance and fairness of transformer-based models. We make the following key contributions. First, we curate a list of Dutch Social Group Terms that reflect social context. Second, we generate counterfactual data for Dutch hate speech using LLMs and established strategies like Manual Group Substitution (MGS) and Sentence Log-Likelihood (SLL). Through qualitative evaluation, we highlight the challenges of generating realistic counterfactuals, particularly with Dutch grammar and contextual coherence. Third, we fine-tune baseline transformer-based models with counterfactual data and evaluate their performance in detecting hate speech. Fourth, we assess the fairness of these models using Counterfactual Token Fairness (CTF) and group fairness metrics, including equality of odds and demographic parity. Our analysis shows that models perform better in terms of hate speech detection, average counterfactual fairness and group fairness. This work addresses a significant gap in the literature on counterfactual fairness for hate speech detection in Dutch and provides practical insights and recommendations for improving both model performance and fairness.
Authors: Sara Rajaram, R. James Cotton, Fabian H. Sinz
Abstract: Preference-based Reinforcement Learning (PbRL) entails a variety of approaches for aligning models with human intent to alleviate the burden of reward engineering. However, most previous PbRL work has not investigated the robustness to labeler errors, inevitable with labelers who are non-experts or operate under time constraints. Additionally, PbRL algorithms often target very specific settings (e.g. pairwise ranked preferences or purely offline learning). We introduce Similarity as Reward Alignment (SARA), a simple contrastive framework that is both resilient to noisy labels and adaptable to diverse feedback formats and training paradigms. SARA learns a latent representation of preferred samples and computes rewards as similarities to the learned latent. We demonstrate strong performance compared to baselines on continuous control offline RL benchmarks. We further demonstrate SARA's versatility in applications such as trajectory filtering for downstream tasks, cross-task preference transfer, and reward shaping in online learning.
Authors: Farida Mohsen, Ali Safa
Abstract: Accurate rotational odometry is crucial for autonomous robotic systems, particularly for small, power-constrained platforms such as drones and mobile robots. This study introduces thermal-gyro fusion, a novel sensor fusion approach that integrates ultra-low-resolution thermal imaging with gyroscope readings for rotational odometry. Unlike RGB cameras, thermal imaging is invariant to lighting conditions and, when fused with gyroscopic data, mitigates drift which is a common limitation of inertial sensors. We first develop a multimodal data acquisition system to collect synchronized thermal and gyroscope data, along with rotational speed labels, across diverse environments. Subsequently, we design and train a lightweight Convolutional Neural Network (CNN) that fuses both modalities for rotational speed estimation. Our analysis demonstrates that thermal-gyro fusion enables a significant reduction in thermal camera resolution without significantly compromising accuracy, thereby improving computational efficiency and memory utilization. These advantages make our approach well-suited for real-time deployment in resource-constrained robotic systems. Finally, to facilitate further research, we publicly release our dataset as supplementary material.
Authors: Xiaoran Fan, Zhichao Sun, Yangfan Gao, Jingfei Xiong, Hang Yan, Yifei Cao, Jiajun Sun, Shuo Li, Zhihao Zhang, Zhiheng Xi, Yuhao Zhou, Senjie Jin, Changhao Jiang, Junjie Ye, Ming Zhang, Rui Zheng, Zhenhua Han, Yunke Zhang, Demei Yan, Shaokang Dong, Tao Ji, Tao Gui, Qi Zhang, Xuanjing Huang
Abstract: Speech-language models (SLMs) offer a promising path toward unifying speech and text understanding and generation. However, challenges remain in achieving effective cross-modal alignment and high-quality speech generation. In this work, we systematically investigate the impact of key components (i.e., speech tokenizers, speech heads, and speaker modeling) on the performance of LLM-centric SLMs. We compare coupled, semi-decoupled, and fully decoupled speech tokenizers under a fair SLM framework and find that decoupled tokenization significantly improves alignment and synthesis quality. To address the information density mismatch between speech and text, we introduce multi-token prediction (MTP) into SLMs, enabling each hidden state to decode multiple speech tokens. This leads to up to 12$\times$ faster decoding and a substantial drop in word error rate (from 6.07 to 3.01). Furthermore, we propose a speaker-aware generation paradigm and introduce RoleTriviaQA, a large-scale role-playing knowledge QA benchmark with diverse speaker identities. Experiments demonstrate that our methods enhance both knowledge understanding and speaker consistency.
Authors: Shuo Yang, Yuqin Dai, Guoqing Wang, Xinran Zheng, Jinfeng Xu, Jinze Li, Zhenzhe Ying, Weiqiang Wang, Edith C. H. Ngai
Abstract: Large Language Models (LLMs) hold significant potential for advancing fact-checking by leveraging their capabilities in reasoning, evidence retrieval, and explanation generation. However, existing benchmarks fail to comprehensively evaluate LLMs and Multimodal Large Language Models (MLLMs) in realistic misinformation scenarios. To bridge this gap, we introduce RealFactBench, a comprehensive benchmark designed to assess the fact-checking capabilities of LLMs and MLLMs across diverse real-world tasks, including Knowledge Validation, Rumor Detection, and Event Verification. RealFactBench consists of 6K high-quality claims drawn from authoritative sources, encompassing multimodal content and diverse domains. Our evaluation framework further introduces the Unknown Rate (UnR) metric, enabling a more nuanced assessment of models' ability to handle uncertainty and balance between over-conservatism and over-confidence. Extensive experiments on 7 representative LLMs and 4 MLLMs reveal their limitations in real-world fact-checking and offer valuable insights for further research. RealFactBench is publicly available at https://github.com/kalendsyang/RealFactBench.git.
Authors: Catalin E. Brita, Hieu Nguyen, Lohithsai Yadala Chanchu, Domonkos Nagy, Maksim Zhdanov
Abstract: Self-attention scales quadratically with input size, limiting its use for large-scale physical systems. Although sparse attention mechanisms provide a viable alternative, they are primarily designed for regular structures such as text or images, making them inapplicable for irregular geometries. In this work, we present Ball Sparse Attention (BSA), which adapts Native Sparse Attention (NSA) (Yuan et al., 2025) to unordered point sets by imposing regularity using the Ball Tree structure from the Erwin Transformer (Zhdanov et al., 2025). We modify NSA's components to work with ball-based neighborhoods, yielding a global receptive field at sub-quadratic cost. On an airflow pressure prediction task, we achieve accuracy comparable to Full Attention while significantly reducing the theoretical computational complexity. Our implementation is available at https://github.com/britacatalin/bsa.
Authors: Ejafa Bassam, Dawei Zhu, Kaigui Bian
Abstract: Knowledge distillation is a model compression technique in which a compact "student" network is trained to replicate the predictive behavior of a larger "teacher" network. In logit-based knowledge distillation it has become the de facto approach to augment cross-entropy with a distillation term. Typically this term is either a KL divergence-matching marginal probabilities or a correlation-based loss capturing intra- and inter-class relationships but in every case it sits as an add-on to cross-entropy with its own weight that must be carefully tuned. In this paper we adopt a choice-theoretic perspective and recast knowledge distillation under the Plackett-Luce model by interpreting teacher logits as "worth" scores. We introduce Plackett-Luce Distillation (PLD), a weighted list-wise ranking loss in which the teacher model transfers knowledge of its full ranking of classes, weighting each ranked choice by its own confidence. PLD directly optimizes a single teacher-optimal ranking of the true label first, followed by the remaining classes in descending teacher confidence, yielding a convex, translation-invariant surrogate that subsumes weighted cross-entropy. Empirically on standard image classification benchmarks, PLD improves Top-1 accuracy by an average of +0.42% over DIST (arXiv:2205.10536) and +1.04% over KD (arXiv:1503.02531) in homogeneous settings and by +0.48% and +1.09% over DIST and KD, respectively, in heterogeneous settings.
Authors: Jingxuan Zhang, Zhenhua Xu, Rui Hu, Wenpeng Xing, Xuhong Zhang, Meng Han
Abstract: Large Language Models (LLMs) have become increasingly prevalent across various sectors, raising critical concerns about model ownership and intellectual property protection. Although backdoor-based fingerprinting has emerged as a promising solution for model authentication, effective attacks for removing these fingerprints remain largely unexplored. Therefore, we present Mismatched Eraser (MEraser), a novel method for effectively removing backdoor-based fingerprints from LLMs while maintaining model performance. Our approach leverages a two-phase fine-tuning strategy utilizing carefully constructed mismatched and clean datasets. Through extensive evaluation across multiple LLM architectures and fingerprinting methods, we demonstrate that MEraser achieves complete fingerprinting removal while maintaining model performance with minimal training data of fewer than 1,000 samples. Furthermore, we introduce a transferable erasure mechanism that enables effective fingerprinting removal across different models without repeated training. In conclusion, our approach provides a practical solution for fingerprinting removal in LLMs, reveals critical vulnerabilities in current fingerprinting techniques, and establishes comprehensive evaluation benchmarks for developing more resilient model protection methods in the future.
Authors: Zain Muhammad Mujahid, Dilshod Azizov, Maha Tufail Agro, Preslav Nakov
Abstract: In an age characterized by the proliferation of mis- and disinformation online, it is critical to empower readers to understand the content they are reading. Important efforts in this direction rely on manual or automatic fact-checking, which can be challenging for emerging claims with limited information. Such scenarios can be handled by assessing the reliability and the political bias of the source of the claim, i.e., characterizing entire news outlets rather than individual claims or articles. This is an important but understudied research direction. While prior work has looked into linguistic and social contexts, we do not analyze individual articles or information in social media. Instead, we propose a novel methodology that emulates the criteria that professional fact-checkers use to assess the factuality and political bias of an entire outlet. Specifically, we design a variety of prompts based on these criteria and elicit responses from large language models (LLMs), which we aggregate to make predictions. In addition to demonstrating sizable improvements over strong baselines via extensive experiments with multiple LLMs, we provide an in-depth error analysis of the effect of media popularity and region on model performance. Further, we conduct an ablation study to highlight the key components of our dataset that contribute to these improvements. To facilitate future research, we released our dataset and code at https://github.com/mbzuai-nlp/llm-media-profiling.
Authors: James E. Smith
Abstract: Active dendrites are the basis for biologically plausible neural networks possessing many desirable features of the biological brain including flexibility, dynamic adaptability, and energy efficiency. A formulation for active dendrites using the notational language of conventional machine learning is put forward as an alternative to a spiking neuron formulation. Based on this formulation, neuromorphic dendrites are developed as basic neural building blocks capable of dynamic online clustering. Features and capabilities of neuromorphic dendrites are demonstrated via a benchmark drawn from experimental neuroscience: spike sorting. Spike sorting takes inputs from electrical probes implanted in neural tissue, detects voltage spikes (action potentials) emitted by neurons, and attempts to sort the spikes according to the neuron that emitted them. Many spike sorting methods form clusters based on the shapes of action potential waveforms, under the assumption that spikes emitted by a given neuron have similar shapes and will therefore map to the same cluster. Using a stream of synthetic spike shapes, the accuracy of the proposed dendrite is compared with the more compute-intensive, offline k-means clustering approach. Overall, the dendrite outperforms k-means and has the advantage of requiring only a single pass through the input stream, learning as it goes. The capabilities of the neuromorphic dendrite are demonstrated for a number of scenarios including dynamic changes in the input stream, differing neuron spike rates, and varying neuron counts.
Authors: Yijun Bian, Lei You
Abstract: Enhancing fairness in machine learning (ML) systems is increasingly important nowadays. While current research focuses on assistant tools for ML pipelines to promote fairness within them, we argue that: 1) The significance of properly defined fairness measures remains underestimated; and 2) Fairness research in ML should integrate societal considerations. The reasons include that detecting discrimination is critical due to the widespread deployment of ML systems and that human-AI feedback loops amplify biases, even when only small social and political biases persist.
Authors: Chunjiang Wang, Kun Zhang, Yandong Liu, Zhiyang He, Xiaodong Tao, S. Kevin Zhou
Abstract: The concept bottleneck model (CBM), as a technique improving interpretability via linking predictions to human-understandable concepts, makes high-risk and life-critical medical image classification credible. Typically, existing CBM methods associate the final layer of visual encoders with concepts to explain the model's predictions. However, we empirically discover the phenomenon of concept preference variation, that is, the concepts are preferably associated with the features at different layers than those only at the final layer; yet a blind last-layer-based association neglects such a preference variation and thus weakens the accurate correspondences between features and concepts, impairing model interpretability. To address this issue, we propose a novel Multi-layer Visual Preference-enhanced Concept Bottleneck Model (MVP-CBM), which comprises two key novel modules: (1) intra-layer concept preference modeling, which captures the preferred association of different concepts with features at various visual layers, and (2) multi-layer concept sparse activation fusion, which sparsely aggregates concept activations from multiple layers to enhance performance. Thus, by explicitly modeling concept preferences, MVP-CBM can comprehensively leverage multi-layer visual information to provide a more nuanced and accurate explanation of model decisions. Extensive experiments on several public medical classification benchmarks demonstrate that MVP-CBM achieves state-of-the-art accuracy and interoperability, verifying its superiority. Code is available at https://github.com/wcj6/MVP-CBM.
Authors: Saksorn Ruangtanusak, Natthapath Rungseesiripak, Peerawat Rojratchadakorn, Monthol Charattrakool, Natapong Nitarach
Abstract: In this paper, we introduce DoTA-RAG (Dynamic-of-Thought Aggregation RAG), a retrieval-augmented generation system optimized for high-throughput, large-scale web knowledge indexes. Traditional RAG pipelines often suffer from high latency and limited accuracy over massive, diverse datasets. DoTA-RAG addresses these challenges with a three-stage pipeline: query rewriting, dynamic routing to specialized sub-indexes, and multi-stage retrieval and ranking. We further enhance retrieval by evaluating and selecting a superior embedding model, re-embedding the large FineWeb-10BT corpus. Moreover, we create a diverse Q&A dataset of 500 questions generated via the DataMorgana setup across a broad range of WebOrganizer topics and formats. DoTA-RAG improves the answer correctness score from 0.752 (baseline, using LiveRAG pre-built vector store) to 1.478 while maintaining low latency, and it achieves a 0.929 correctness score on the Live Challenge Day. These results highlight DoTA-RAG's potential for practical deployment in domains requiring fast, reliable access to large and evolving knowledge sources.
Authors: Ananya Joshi, Celia Cintas, Skyler Speakman
Abstract: Recent work shows that Sparse Autoencoders (SAE) applied to large language model (LLM) layers have neurons corresponding to interpretable concepts. These SAE neurons can be modified to align generated outputs, but only towards pre-identified topics and with some parameter tuning. Our approach leverages the observational and modification properties of SAEs to enable alignment for any topic. This method 1) scores each SAE neuron by its semantic similarity to an alignment text and uses them to 2) modify SAE-layer-level outputs by emphasizing topic-aligned neurons. We assess the alignment capabilities of this approach on diverse public topic datasets including Amazon reviews, Medicine, and Sycophancy, across the currently available open-source LLMs and SAE pairs (GPT2 and Gemma) with multiple SAEs configurations. Experiments aligning to medical prompts reveal several benefits over fine-tuning, including increased average language acceptability (0.25 vs. 0.5), reduced training time across multiple alignment topics (333.6s vs. 62s), and acceptable inference time for many applications (+0.00092s/token). Our open-source code is available at github.com/IBM/sae-steering.
Authors: Jie Pan, Tianyi Wang, Christian Claudel, Jing Shi
Abstract: Intelligent transportation systems require connected and automated vehicles (CAVs) to conduct safe and efficient cooperation with human-driven vehicles (HVs) in complex real-world traffic environments. However, the inherent unpredictability of human behaviour, especially at bottlenecks such as highway on-ramp merging areas, often disrupts traffic flow and compromises system performance. To address the challenge of cooperative on-ramp merging in heterogeneous traffic environments, this study proposes a trust-based multi-agent reinforcement learning (Trust-MARL) framework. At the macro level, Trust-MARL enhances global traffic efficiency by leveraging inter-agent trust to improve bottleneck throughput and mitigate traffic shockwave through emergent group-level coordination. At the micro level, a dynamic trust mechanism is designed to enable CAVs to adjust their cooperative strategies in response to real-time behaviors and historical interactions with both HVs and other CAVs. Furthermore, a trust-triggered game-theoretic decision-making module is integrated to guide each CAV in adapting its cooperation factor and executing context-aware lane-changing decisions under safety, comfort, and efficiency constraints. An extensive set of ablation studies and comparative experiments validates the effectiveness of the proposed Trust-MARL approach, demonstrating significant improvements in safety, efficiency, comfort, and adaptability across varying CAV penetration rates and traffic densities.
Authors: Tzu-Quan Lin, Heng-Cheng Kuo, Tzu-Chieh Wei, Hsi-Chun Cheng, Chun-Wei Chen, Hsien-Fu Hsiao, Yu Tsao, Hung-yi Lee
Abstract: While Mamba has demonstrated strong performance in language modeling, its potential as a speech self-supervised (SSL) model remains underexplored, with prior studies limited to isolated tasks. To address this, we explore Mamba-based HuBERT models as alternatives to Transformer-based SSL architectures. Leveraging the linear-time Selective State Space, these models enable fine-tuning on long-context ASR with significantly lower compute. Moreover, they show superior performance when fine-tuned for streaming ASR. Beyond fine-tuning, these models show competitive performance on SUPERB probing benchmarks, particularly in causal settings. Our analysis shows that they yield higher-quality quantized representations and capture speaker-related features more distinctly than Transformer-based models. These findings highlight Mamba-based SSL as a promising and complementary direction for long-sequence modeling, real-time speech modeling, and speech unit extraction.
Authors: Nagham Hamad, Mohammed Khalilia, Mustafa Jarrar
Abstract: We introduce Konooz, a novel multi-dimensional corpus covering 16 Arabic dialects across 10 domains, resulting in 160 distinct corpora. The corpus comprises about 777k tokens, carefully collected and manually annotated with 21 entity types using both nested and flat annotation schemes - using the Wojood guidelines. While Konooz is useful for various NLP tasks like domain adaptation and transfer learning, this paper primarily focuses on benchmarking existing Arabic Named Entity Recognition (NER) models, especially cross-domain and cross-dialect model performance. Our benchmarking of four Arabic NER models using Konooz reveals a significant drop in performance of up to 38% when compared to the in-distribution data. Furthermore, we present an in-depth analysis of domain and dialect divergence and the impact of resource scarcity. We also measured the overlap between domains and dialects using the Maximum Mean Discrepancy (MMD) metric, and illustrated why certain NER models perform better on specific dialects and domains. Konooz is open-source and publicly available at https://sina.birzeit.edu/wojood/#download
Authors: Mingxuan Cui, Duo Zhou, Yuxuan Han, Grani A. Hanasusanto, Qiong Wang, Huan Zhang, Zhengyuan Zhou
Abstract: Deep reinforcement learning (RL) has achieved significant success, yet its application in real-world scenarios is often hindered by a lack of robustness to environmental uncertainties. To solve this challenge, some robust RL algorithms have been proposed, but most are limited to tabular settings. In this work, we propose Distributionally Robust Soft Actor-Critic (DR-SAC), a novel algorithm designed to enhance the robustness of the state-of-the-art Soft Actor-Critic (SAC) algorithm. DR-SAC aims to maximize the expected value with entropy against the worst possible transition model lying in an uncertainty set. A distributionally robust version of the soft policy iteration is derived with a convergence guarantee. For settings where nominal distributions are unknown, such as offline RL, a generative modeling approach is proposed to estimate the required nominal distributions from data. Furthermore, experimental results on a range of continuous control benchmark tasks demonstrate our algorithm achieves up to $9.8$ times the average reward of the SAC baseline under common perturbations. Additionally, compared with existing robust reinforcement learning algorithms, DR-SAC significantly improves computing efficiency and applicability to large-scale problems.
Authors: Valentin Ackva, Fares Schulz
Abstract: Numerous tools for neural network inference are currently available, yet many do not meet the requirements of real-time audio applications. In response, we introduce anira, an efficient cross-platform library. To ensure compatibility with a broad range of neural network architectures and frameworks, anira supports ONNX Runtime, LibTorch, and TensorFlow Lite as backends. Each inference engine exhibits real-time violations, which anira mitigates by decoupling the inference from the audio callback to a static thread pool. The library incorporates built-in latency management and extensive benchmarking capabilities, both crucial to ensure a continuous signal flow. Three different neural network architectures for audio effect emulation are then subjected to benchmarking across various configurations. Statistical modeling is employed to identify the influence of various factors on performance. The findings indicate that for stateless models, ONNX Runtime exhibits the lowest runtimes. For stateful models, LibTorch demonstrates the fastest performance. Our results also indicate that for certain model-engine combinations, the initial inferences take longer, particularly when these inferences exhibit a higher incidence of real-time violations.
Authors: Bilal Saleh Husain
Abstract: Large Language Models (LLMs) have demonstrated remarkable capabilities, yet their susceptibility to adversarial attacks, particularly jailbreaking, poses significant safety and ethical concerns. While numerous jailbreak methods exist, many suffer from computational expense, high token usage, or complex decoding schemes. Liu et al. (2024) introduced FlipAttack, a black-box method that achieves high attack success rates (ASR) through simple prompt manipulation. This paper investigates the underlying mechanisms of FlipAttack's effectiveness by analyzing the semantic changes induced by its flipping modes. We hypothesize that semantic dissimilarity between original and manipulated prompts is inversely correlated with ASR. To test this, we examine embedding space visualizations (UMAP, KDE) and cosine similarities for FlipAttack's modes. Furthermore, we introduce a novel adversarial attack, Alphabet Index Mapping (AIM), designed to maximize semantic dissimilarity while maintaining simple decodability. Experiments on GPT-4 using a subset of AdvBench show AIM and its variant AIM+FWO achieve a 94% ASR, outperforming FlipAttack and other methods on this subset. Our findings suggest that while high semantic dissimilarity is crucial, a balance with decoding simplicity is key for successful jailbreaking. This work contributes to a deeper understanding of adversarial prompt mechanics and offers a new, effective jailbreak technique.
Authors: Bianca Trinkenreich, Fabio Calefato, Geir Hanssen, Kelly Blincoe, Marcos Kalinowski, Mauro Pezz\`e, Paolo Tell, Margaret-Anne Storey
Abstract: The adoption of Large Language Models (LLMs) is not only transforming software engineering (SE) practice but is also poised to fundamentally disrupt how research is conducted in the field. While perspectives on this transformation range from viewing LLMs as mere productivity tools to considering them revolutionary forces, we argue that the SE research community must proactively engage with and shape the integration of LLMs into research practices, emphasizing human agency in this transformation. As LLMs rapidly become integral to SE research - both as tools that support investigations and as subjects of study - a human-centric perspective is essential. Ensuring human oversight and interpretability is necessary for upholding scientific rigor, fostering ethical responsibility, and driving advancements in the field. Drawing from discussions at the 2nd Copenhagen Symposium on Human-Centered AI in SE, this position paper employs McLuhan's Tetrad of Media Laws to analyze the impact of LLMs on SE research. Through this theoretical lens, we examine how LLMs enhance research capabilities through accelerated ideation and automated processes, make some traditional research practices obsolete, retrieve valuable aspects of historical research approaches, and risk reversal effects when taken to extremes. Our analysis reveals opportunities for innovation and potential pitfalls that require careful consideration. We conclude with a call to action for the SE research community to proactively harness the benefits of LLMs while developing frameworks and guidelines to mitigate their risks, to ensure continued rigor and impact of research in an AI-augmented future.
Authors: Yuxiang Wang, Xuecheng Bai, Boyu Hu, Chuanzhi Xu, Haodong Chen, Vera Chung, Tingxue Li
Abstract: Small object detection in UAV imagery is crucial for applications such as search-and-rescue, traffic monitoring, and environmental surveillance, but it is hampered by tiny object size, low signal-to-noise ratios, and limited feature extraction. Existing multi-scale fusion methods help, but add computational burden and blur fine details, making small object detection in cluttered scenes difficult. To overcome these challenges, we propose the Multi-scale Global-detail Feature Integration Strategy (MGDFIS), a unified fusion framework that tightly couples global context with local detail to boost detection performance while maintaining efficiency. MGDFIS comprises three synergistic modules: the FusionLock-TSS Attention Module, which marries token-statistics self-attention with DynamicTanh normalization to highlight spectral and spatial cues at minimal cost; the Global-detail Integration Module, which fuses multi-scale context via directional convolution and parallel attention while preserving subtle shape and texture variations; and the Dynamic Pixel Attention Module, which generates pixel-wise weighting maps to rebalance uneven foreground and background distributions and sharpen responses to true object regions. Extensive experiments on the VisDrone benchmark demonstrate that MGDFIS consistently outperforms state-of-the-art methods across diverse backbone architectures and detection frameworks, achieving superior precision and recall with low inference time. By striking an optimal balance between accuracy and resource usage, MGDFIS provides a practical solution for small-object detection on resource-constrained UAV platforms.
Authors: Cuong Manh Hoang, Yeejin Lee, Byeongkeun Kang
Abstract: This work addresses the task of self-supervised learning (SSL) on a long-tailed dataset that aims to learn balanced and well-separated representations for downstream tasks such as image classification. This task is crucial because the real world contains numerous object categories, and their distributions are inherently imbalanced. Towards robust SSL on a class-imbalanced dataset, we investigate leveraging a network trained using unlabeled out-of-distribution (OOD) data that are prevalently available online. We first train a network using both in-domain (ID) and sampled OOD data by back-propagating the proposed pseudo semantic discrimination loss alongside a domain discrimination loss. The OOD data sampling and loss functions are designed to learn a balanced and well-separated embedding space. Subsequently, we further optimize the network on ID data by unsupervised contrastive learning while using the previously trained network as a guiding network. The guiding network is utilized to select positive/negative samples and to control the strengths of attractive/repulsive forces in contrastive learning. We also distil and transfer its embedding space to the training network to maintain balancedness and separability. Through experiments on four publicly available long-tailed datasets, we demonstrate that the proposed method outperforms previous state-of-the-art methods.
Authors: Wenhong Zhu, Ruobing Xie, Weinan Zhang, Rui Wang
Abstract: Realignment becomes necessary when a language model (LM) fails to meet expected performance. We propose a flexible realignment framework that supports quantitative control of alignment degree during training and inference. This framework incorporates Training-time Realignment (TrRa), which efficiently realigns the reference model by leveraging the controllable fusion of logits from both the reference and already aligned models. For example, TrRa reduces token usage by 54.63% on DeepSeek-R1-Distill-Qwen-1.5B without any performance degradation, outperforming DeepScaleR-1.5B's 33.86%. To complement TrRa during inference, we introduce a layer adapter that enables smooth Inference-time Realignment (InRa). This adapter is initialized to perform an identity transformation at the bottom layer and is inserted preceding the original layers. During inference, input embeddings are simultaneously processed by the adapter and the original layer, followed by the remaining layers, and then controllably interpolated at the logit level. We upgraded DeepSeek-R1-Distill-Qwen-7B from a slow-thinking model to one that supports both fast and slow thinking, allowing flexible alignment control even during inference. By encouraging deeper reasoning, it even surpassed its original performance.
Authors: Jiaming Zhang, Xin Wang, Xingjun Ma, Lingyu Qiu, Yu-Gang Jiang, Jitao Sang
Abstract: Vision-Language Models (VLMs) such as CLIP have demonstrated remarkable capabilities in understanding relationships between visual and textual data through joint embedding spaces. Despite their effectiveness, these models remain vulnerable to adversarial attacks, particularly in the image modality, posing significant security concerns. Building upon our previous work on Adversarial Prompt Tuning (AdvPT), which introduced learnable text prompts to enhance adversarial robustness in VLMs without extensive parameter training, we present a significant extension by introducing the Neural Augmentor framework for Multi-modal Adversarial Prompt Tuning (NAP-Tuning).Our key innovations include: (1) extending AdvPT from text-only to multi-modal prompting across both text and visual modalities, (2) expanding from single-layer to multi-layer prompt architectures, and (3) proposing a novel architecture-level redesign through our Neural Augmentor approach, which implements feature purification to directly address the distortions introduced by adversarial attacks in feature space. Our NAP-Tuning approach incorporates token refiners that learn to reconstruct purified features through residual connections, allowing for modality-specific and layer-specific feature correction.Comprehensive experiments demonstrate that NAP-Tuning significantly outperforms existing methods across various datasets and attack types. Notably, our approach shows significant improvements over the strongest baselines under the challenging AutoAttack benchmark, outperforming them by 33.5% on ViT-B16 and 33.0% on ViT-B32 architectures while maintaining competitive clean accuracy.
Authors: Pengfei Zuo, Huimin Lin, Junbo Deng, Nan Zou, Xingkun Yang, Yingyu Diao, Weifeng Gao, Ke Xu, Zhangyu Chen, Shirui Lu, Zhao Qiu, Peiyang Li, Xianyu Chang, Zhengzhong Yu, Fangzheng Miao, Jia Zheng, Ying Li, Yuan Feng, Bei Wang, Zaijian Zong, Mosong Zhou, Wenli Zhou, Houjiang Chen, Xingyu Liao, Yipeng Li, Wenxiao Zhang, Ping Zhu, Yinggang Wang, Chuanjie Xiao, Depeng Liang, Dong Cao, Juncheng Liu, Yongqiang Yang, Xiaolong Bai, Yi Li, Huaguo Xie, Huatao Wu, Zhibin Yu, Lv Chen, Hu Liu, Yujun Ding, Haipei Zhu, Jing Xia, Yi Xiong, Zhou Yu, Heng Liao
Abstract: The rapid evolution of large language models (LLMs), driven by growing parameter scales, adoption of mixture-of-experts (MoE) architectures, and expanding context lengths, imposes unprecedented demands on AI infrastructure. Traditional AI clusters face limitations in compute intensity, memory bandwidth, inter-chip communication, and latency, compounded by variable workloads and strict service-level objectives. Addressing these issues requires fundamentally redesigned hardware-software integration. This paper introduces Huawei CloudMatrix, a next-generation AI datacenter architecture, realized in the production-grade CloudMatrix384 supernode. It integrates 384 Ascend 910C NPUs and 192 Kunpeng CPUs interconnected via an ultra-high-bandwidth Unified Bus (UB) network, enabling direct all-to-all communication and dynamic pooling of resources. These features optimize performance for communication-intensive operations, such as large-scale MoE expert parallelism and distributed key-value cache access. To fully leverage CloudMatrix384, we propose CloudMatrix-Infer, an advanced LLM serving solution incorporating three core innovations: a peer-to-peer serving architecture that independently scales prefill, decode, and caching; a large-scale expert parallelism strategy supporting EP320 via efficient UB-based token dispatch; and hardware-aware optimizations including specialized operators, microbatch-based pipelining, and INT8 quantization. Evaluation with the DeepSeek-R1 model shows CloudMatrix-Infer achieves state-of-the-art efficiency: prefill throughput of 6,688 tokens/s per NPU and decode throughput of 1,943 tokens/s per NPU (<50 ms TPOT). It effectively balances throughput and latency, sustaining 538 tokens/s even under stringent 15 ms latency constraints, while INT8 quantization maintains model accuracy across benchmarks.
Authors: Ye Li, Yuan Meng, Zewen Sun, Kangye Ji, Chen Tang, Jiajun Fan, Xinzhu Ma, Shutao Xia, Zhi Wang, Wenwu Zhu
Abstract: Vision-Language-Action (VLA) models have attracted increasing attention for their strong control capabilities. However, their high computational cost and low execution frequency hinder their suitability for real-time tasks such as robotic manipulation and autonomous navigation. Existing VLA acceleration methods primarily focus on structural optimization, overlooking the fact that these models operate in sequential decision-making environments. As a result, temporal redundancy in sequential action generation and spatial redundancy in visual input remain unaddressed. To this end, we propose SP-VLA, a unified framework that accelerates VLA models by jointly scheduling models and pruning tokens. Specifically, we design an action-aware model scheduling mechanism that reduces temporal redundancy by dynamically switching between VLA model and a lightweight generator. Inspired by the human motion pattern of focusing on key decision points while relying on intuition for other actions, we categorize VLA actions into deliberative and intuitive, assigning the former to the VLA model and the latter to the lightweight generator, enabling frequency-adaptive execution through collaborative model scheduling. To address spatial redundancy, we further develop a spatio-semantic dual-aware token pruning method. Tokens are classified into spatial and semantic types and pruned based on their dual-aware importance to accelerate VLA inference. These two mechanisms work jointly to guide the VLA in focusing on critical actions and salient visual information, achieving effective acceleration while maintaining high accuracy. Experimental results demonstrate that our method achieves up to 1.5$\times$ acceleration with less than 3% drop in accuracy, outperforming existing approaches in multiple tasks.
Authors: Zhilin Lin, Shiliang Sun
Abstract: Reinforcement learning (RL) is playing an increasingly important role in fields such as robotic control and autonomous driving. However, the gap between simulation and the real environment remains a major obstacle to the practical deployment of RL. Agents trained in simulators often struggle to maintain performance when transferred to real-world physical environments. In this paper, we propose a latent space based approach to analyze the impact of simulation on real-world policy improvement in model-based settings. As a natural extension of model-based methods, our approach enables an intuitive observation of the challenges faced by model-based methods in sim-to-real transfer. Experiments conducted in the MuJoCo environment evaluate the performance of our method in both measuring and mitigating the sim-to-real gap. The experiments also highlight the various challenges that remain in overcoming the sim-to-real gap, especially for model-based methods.
Authors: Hang Xu, Wei Yu, Jiangtong Tan, Zhen Zou, Feng Zhao
Abstract: Blind Super-Resolution (blind SR) aims to enhance the model's generalization ability with unknown degradation, yet it still encounters severe overfitting issues. Some previous methods inspired by dropout, which enhances generalization by regularizing features, have shown promising results in blind SR. Nevertheless, these methods focus solely on regularizing features before the final layer and overlook the need for generalization in features at intermediate layers. Without explicit regularization of features at intermediate layers, the blind SR network struggles to obtain well-generalized feature representations. However, the key challenge is that directly applying dropout to intermediate layers leads to a significant performance drop, which we attribute to the inconsistency in training-testing and across layers it introduced. Therefore, we propose Adaptive Dropout, a new regularization method for blind SR models, which mitigates the inconsistency and facilitates application across intermediate layers of networks. Specifically, for training-testing inconsistency, we re-design the form of dropout and integrate the features before and after dropout adaptively. For inconsistency in generalization requirements across different layers, we innovatively design an adaptive training strategy to strengthen feature propagation by layer-wise annealing. Experimental results show that our method outperforms all past regularization methods on both synthetic and real-world benchmark datasets, also highly effective in other image restoration tasks. Code is available at \href{https://github.com/xuhang07/Adpative-Dropout}{https://github.com/xuhang07/Adpative-Dropout}.
URLs: https://github.com/xuhang07/Adpative-Dropout, https://github.com/xuhang07/Adpative-Dropout
Authors: Rong Wu, Ziqi Chen, Liming Zhong, Heng Li, Hai Shu
Abstract: Existing segmentation models trained on a single medical imaging dataset often lack robustness when encountering unseen organs or tumors. Developing a robust model capable of identifying rare or novel tumor categories not present during training is crucial for advancing medical imaging applications. We propose DSM, a novel framework that leverages diffusion and state space models to segment unseen tumor categories beyond the training data. DSM utilizes two sets of object queries trained within modified attention decoders to enhance classification accuracy. Initially, the model learns organ queries using an object-aware feature grouping strategy to capture organ-level visual features. It then refines tumor queries by focusing on diffusion-based visual prompts, enabling precise segmentation of previously unseen tumors. Furthermore, we incorporate diffusion-guided feature fusion to improve semantic segmentation performance. By integrating CLIP text embeddings, DSM captures category-sensitive classes to improve linguistic transfer knowledge, thereby enhancing the model's robustness across diverse scenarios and multi-label tasks. Extensive experiments demonstrate the superior performance of DSM in various tumor segmentation tasks. Code is available at https://github.com/Rows21/KMax-Mamba.
Authors: Chaoyi Lu, Yiding Sun, Jinqian Chen, Zhichuan Yang, Jiangming Pan, Jihua Zhu
Abstract: Asynchronous federated learning (AFL) accelerates training by eliminating the need to wait for stragglers, but its asynchronous nature introduces gradient staleness, where outdated gradients degrade performance. Existing solutions address this issue with gradient buffers, forming a semi-asynchronous framework. However, this approach struggles when buffers accumulate numerous stale gradients, as blindly aggregating all gradients can harm training. To address this, we propose AFBS (Asynchronous FL Buffer Selection), the first algorithm to perform gradient selection within buffers while ensuring privacy protection. Specifically, the client sends the random projection encrypted label distribution matrix before training, and the server performs client clustering based on it. During training, server scores and selects gradients within each cluster based on their informational value, discarding low-value gradients to enhance semi-asynchronous federated learning. Extensive experiments in highly heterogeneous system and data environments demonstrate AFBS's superior performance compared to state-of-the-art methods. Notably, on the most challenging task, CIFAR-100, AFBS improves accuracy by up to 4.8% over the previous best algorithm and reduces the time to reach target accuracy by 75%.
Authors: Adrian Rubio-Solis, Luciano Nava-Balanzar, Tomas Salgado-Jimenez
Abstract: In autonomous underwater missions, the successful completion of predefined paths mainly depends on the ability of underwater vehicles to recognise their surroundings. In this study, we apply the concept of Fast Interval Type-2 Fuzzy Extreme Learning Machine (FIT2-FELM) to train a Takagi-Sugeno-Kang IT2 Fuzzy Inference System (TSK IT2-FIS) for on-board sonar data classification using an underwater vehicle called BlueROV2. The TSK IT2-FIS is integrated into a Hierarchical Navigation Strategy (HNS) as the main navigation engine to infer local motions and provide the BlueROV2 with full autonomy to follow an obstacle-free trajectory in a water container of 2.5m x 2.5m x 3.5m. Compared to traditional navigation architectures, using the proposed method, we observe a robust path following behaviour in the presence of uncertainty and noise. We found that the proposed approach provides the BlueROV with a more complete sensory picture about its surroundings while real-time navigation planning is performed by the concurrent execution of two or more tasks.
Authors: Manas Pandey, Bharath Hebbe Madhusudhana, Saikat Ghosh, Dmitry Budker
Abstract: The capabilities of modern artificial intelligence (AI) as a ``scientific collaborator'' are explored by engaging it with three nuanced problems in quantum optics: state populations in optical pumping, resonant transitions between decaying states (the Burshtein effect), and degenerate mirrorless lasing. Through iterative dialogue, the authors observe that AI models--when prompted and corrected--can reason through complex scenarios, refine their answers, and provide expert-level guidance, closely resembling the interaction with an adept colleague. The findings highlight that AI democratizes access to sophisticated modeling and analysis, shifting the focus in scientific practice from technical mastery to the generation and testing of ideas, and reducing the time for completing research tasks from days to minutes.
Authors: Han Ke, Xiao Ke, Ye Yan, Rui Liu, Jinpeng Yang, Tianwen Zhang, Xu Zhan, Xiaowo Xu
Abstract: DL based Synthetic Aperture Radar (SAR) ship detection has tremendous advantages in numerous areas. However, it still faces some problems, such as the lack of prior knowledge, which seriously affects detection accuracy. In order to solve this problem, we propose a scene-aware SAR ship detection method based on unsupervised sea-land segmentation. This method follows a classical two-stage framework and is enhanced by two models: the unsupervised land and sea segmentation module (ULSM) and the land attention suppression module (LASM). ULSM and LASM can adaptively guide the network to reduce attention on land according to the type of scenes (inshore scene and offshore scene) and add prior knowledge (sea land segmentation information) to the network, thereby reducing the network's attention to land directly and enhancing offshore detection performance relatively. This increases the accuracy of ship detection and enhances the interpretability of the model. Specifically, in consideration of the lack of land sea segmentation labels in existing deep learning-based SAR ship detection datasets, ULSM uses an unsupervised approach to classify the input data scene into inshore and offshore types and performs sea-land segmentation for inshore scenes. LASM uses the sea-land segmentation information as prior knowledge to reduce the network's attention to land. We conducted our experiments using the publicly available SSDD dataset, which demonstrated the effectiveness of our network.
Authors: Mehdi Bennis
Abstract: Just like power, water and transportation systems, wireless networks are a crucial societal infrastructure. As natural and human-induced disruptions continue to grow, wireless networks must be resilient to unforeseen events, able to withstand and recover from unexpected adverse conditions, shocks, unmodeled disturbances and cascading failures. Despite its critical importance, resilience remains an elusive concept, with its mathematical foundations still underdeveloped. Unlike robustness and reliability, resilience is premised on the fact that disruptions will inevitably happen. Resilience, in terms of elasticity, focuses on the ability to bounce back to favorable states, while resilience as plasticity involves agents (or networks) that can flexibly expand their states, hypotheses and course of actions, by transforming through real-time adaptation and reconfiguration. This constant situational awareness and vigilance of adapting world models and counterfactually reasoning about potential system failures and the corresponding best responses, is a core aspect of resilience. This article seeks to first define resilience and disambiguate it from reliability and robustness, before delving into the mathematics of resilience. Finally, the article concludes by presenting nuanced metrics and discussing trade-offs tailored to the unique characteristics of network resilience.
Authors: Lei Lv, Yunfei Li, Yu Luo, Fuchun Sun, Tao Kong, Jiafeng Xu, Xiao Ma
Abstract: We present \textbf{FlowRL}, a novel framework for online reinforcement learning that integrates flow-based policy representation with Wasserstein-2-regularized optimization. We argue that in addition to training signals, enhancing the expressiveness of the policy class is crucial for the performance gains in RL. Flow-based generative models offer such potential, excelling at capturing complex, multimodal action distributions. However, their direct application in online RL is challenging due to a fundamental objective mismatch: standard flow training optimizes for static data imitation, while RL requires value-based policy optimization through a dynamic buffer, leading to difficult optimization landscapes. FlowRL first models policies via a state-dependent velocity field, generating actions through deterministic ODE integration from noise. We derive a constrained policy search objective that jointly maximizes Q through the flow policy while bounding the Wasserstein-2 distance to a behavior-optimal policy implicitly derived from the replay buffer. This formulation effectively aligns the flow optimization with the RL objective, enabling efficient and value-aware policy learning despite the complexity of the policy class. Empirical evaluations on DMControl and Humanoidbench demonstrate that FlowRL achieves competitive performance in online reinforcement learning benchmarks.
Authors: David Sweet, Siddhant anand Jadhav
Abstract: Bayesian optimization (BO) has traditionally solved black box problems where evaluation is expensive and, therefore, design-evaluation pairs (i.e., observations) are few. Recently, there has been growing interest in applying BO to problems where evaluation is cheaper and, thus, observations are more plentiful. An impediment to scaling BO to many observations, $N$, is the $O(N^3)$ scaling of a na{\"i}ve query of the Gaussian process (GP) surrogate. Modern implementations reduce this to $O(N^2)$, but the GP remains a bottleneck. We propose Epistemic Nearest Neighbors (ENN), a surrogate that estimates function values and epistemic uncertainty from $K$ nearest-neighbor observations. ENN has $O(N)$ query time and omits hyperparameter fitting, leaving uncertainty uncalibrated. To accommodate the lack of calibration, we employ an acquisition method based on Pareto-optimal tradeoffs between predicted value and uncertainty. Our proposed method, TuRBO-ENN, replaces the GP surrogate in TuRBO with ENN and its Thompson sampling acquisition method with our Pareto-based alternative. We demonstrate numerically that TuRBO-ENN can reduce the time to generate proposals by one to two orders of magnitude compared to TuRBO and scales to thousands of observations.
Authors: Zonghui Yang, Shijian Gao, Xiang Cheng, Liuqing Yang
Abstract: Integrated sensing and communication (ISAC) within sub-THz frequencies is crucial for future air-ground networks, but unique propagation characteristics and hardware limitations present challenges in optimizing ISAC performance while increasing operational latency. This paper introduces a multi-modal sensing fusion framework inspired by synesthesia of machine (SoM) to enhance sub-THz ISAC transmission. By exploiting inherent degrees of freedom in sub-THz hardware and channels, the framework optimizes the radio-frequency environment. Squint-aware beam management is developed to improve air-ground network adaptability, enabling three-dimensional dynamic ISAC links. Leveraging multi-modal information, the framework enhances ISAC performance and reduces latency. Visual data rapidly localizes users and targets, while a customized multi-modal learning algorithm optimizes the hybrid precoder. A new metric provides comprehensive performance evaluation, and extensive experiments demonstrate that the proposed scheme significantly improves ISAC efficiency.
Authors: Jihu Lee, Kunwoong Kim, Yongdai Kim
Abstract: Fair clustering has become a socially significant task with the advancement of machine learning technologies and the growing demand for trustworthy AI. Group fairness ensures that the proportions of each sensitive group are similar in all clusters. Most existing group-fair clustering methods are based on the $K$-means clustering and thus require the distance between instances and the number of clusters to be given in advance. To resolve this limitation, we propose a fair Bayesian model-based clustering called Fair Bayesian Clustering (FBC). We develop a specially designed prior which puts its mass only on fair clusters, and implement an efficient MCMC algorithm. Advantages of FBC are that it can infer the number of clusters and can be applied to any data type as long as the likelihood is defined (e.g., categorical data). Experiments on real-world datasets show that FBC (i) reasonably infers the number of clusters, (ii) achieves a competitive utility-fairness trade-off compared to existing fair clustering methods, and (iii) performs well on categorical data.
Authors: Nina Cai, Jinguang Han
Abstract: Federated learning is a promising distributed learning paradigm that enables collaborative model training without exposing local client data, thereby protect data privacy. However, it also brings new threats and challenges. The advancement of model inversion attacks has rendered the plaintext transmission of local models insecure, while the distributed nature of federated learning makes it particularly vulnerable to attacks raised by malicious clients. To protect data privacy and prevent malicious client attacks, this paper proposes a privacy-preserving federated learning framework based on verifiable functional encryption, without a non-colluding dual-server setup or additional trusted third-party. Specifically, we propose a novel decentralized verifiable functional encryption (DVFE) scheme that enables the verification of specific relationships over multi-dimensional ciphertexts. This scheme is formally treated, in terms of definition, security model and security proof. Furthermore, based on the proposed DVFE scheme, we design a privacy-preserving federated learning framework VFEFL that incorporates a novel robust aggregation rule to detect malicious clients, enabling the effective training of high-accuracy models under adversarial settings. Finally, we provide formal analysis and empirical evaluation of the proposed schemes. The results demonstrate that our approach achieves the desired privacy protection, robustness, verifiability and fidelity, while eliminating the reliance on non-colluding dual-server settings or trusted third parties required by existing methods.
Authors: Weiji Xie, Jinrui Han, Jiakun Zheng, Huanyu Li, Xinzhe Liu, Jiyuan Shi, Weinan Zhang, Chenjia Bai, Xuelong Li
Abstract: Humanoid robots are promising to acquire various skills by imitating human behaviors. However, existing algorithms are only capable of tracking smooth, low-speed human motions, even with delicate reward and curriculum design. This paper presents a physics-based humanoid control framework, aiming to master highly-dynamic human behaviors such as Kungfu and dancing through multi-steps motion processing and adaptive motion tracking. For motion processing, we design a pipeline to extract, filter out, correct, and retarget motions, while ensuring compliance with physical constraints to the maximum extent. For motion imitation, we formulate a bi-level optimization problem to dynamically adjust the tracking accuracy tolerance based on the current tracking error, creating an adaptive curriculum mechanism. We further construct an asymmetric actor-critic framework for policy training. In experiments, we train whole-body control policies to imitate a set of highly-dynamic motions. Our method achieves significantly lower tracking errors than existing approaches and is successfully deployed on the Unitree G1 robot, demonstrating stable and expressive behaviors. The project page is https://kungfu-bot.github.io.
Authors: Frederic Gmeiner, Kaitao Luo, Ye Wang, Kenneth Holstein, Nikolas Martelaro
Abstract: Despite the potential of generative AI (GenAI) design tools to enhance design processes, professionals often struggle to integrate AI into their workflows. Fundamental cognitive challenges include the need to specify all design criteria as distinct parameters upfront (intent formulation) and designers' reduced cognitive involvement in the design process due to cognitive offloading, which can lead to insufficient problem exploration, underspecification, and limited ability to evaluate outcomes. Motivated by these challenges, we envision novel metacognitive support agents that assist designers in working more reflectively with GenAI. To explore this vision, we conducted exploratory prototyping through a Wizard of Oz elicitation study with 20 mechanical designers probing multiple metacognitive support strategies. We found that agent-supported users created more feasible designs than non-supported users, with differing impacts between support strategies. Based on these findings, we discuss opportunities and tradeoffs of metacognitive support agents and considerations for future AI-based design tools.
Authors: Yingru Li
Abstract: We analyzes the logit dynamics of softmax policy gradient methods. We derive the exact formula for the L2 norm of the logit update vector: $$ \|\Delta \mathbf{z}\|_2 \propto \sqrt{1-2P_c + C(P)} $$ This equation demonstrates that update magnitudes are determined by the chosen action's probability ($P_c$) and the policy's collision probability ($C(P)$), a measure of concentration inversely related to entropy. Our analysis reveals an inherent self-regulation mechanism where learning vigor is automatically modulated by policy confidence, providing a foundational insight into the stability and convergence of these methods.
Authors: Erica Cai, Xi Chen, Reagan Grey Keeney, Ethan Zuckerman, Brendan O'Connor, Przemyslaw A. Grabowicz
Abstract: Comparative studies of news coverage are challenging to conduct because methods to identify news articles about the same event in different languages require expertise that is difficult to scale. We introduce an AI-powered method for identifying news articles based on an event FINGERPRINT, which is a minimal set of metadata required to identify critical events. Our event coverage identification method, FINGERPRINT TO ARTICLE MATCHING FOR EVENTS (FAME), efficiently identifies news articles about critical world events, specifically terrorist attacks and several types of natural disasters. FAME does not require training data and is able to automatically and efficiently identify news articles that discuss an event given its fingerprint: time, location, and class (such as storm or flood). The method achieves state-of-the-art performance and scales to massive databases of tens of millions of news articles and hundreds of events happening globally. We use FAME to identify 27,441 articles that cover 470 natural disaster and terrorist attack events that happened in 2020. To this end, we use a massive database of news articles in three languages from MediaCloud, and three widely used, expert-curated databases of critical events: EM-DAT, USGS, and GTD. Our case study reveals patterns consistent with prior literature: coverage of disasters and terrorist attacks correlates to death counts, to the GDP of a country where the event occurs, and to trade volume between the reporting country and the country where the event occurred. We share our NLP annotations and cross-country media attention data to support the efforts of researchers and media monitoring organizations.
Authors: Antonin Sulc, Thorsten Hellert, Aaron Reed, Adam Carpenter, Alex Bien, Chris Tennant, Claudio Bisegni, Daniel Lersch, Daniel Ratner, David Lawrence, Diana McSpadden, Hayden Hoschouer, Jason St. John, Thomas Britton
Abstract: This work demonstrates electronic logbook (eLog) systems leveraging modern AI-driven information retrieval capabilities at the accelerator facilities of Fermilab, Jefferson Lab, Lawrence Berkeley National Laboratory (LBNL), SLAC National Accelerator Laboratory. We evaluate contemporary tools and methodologies for information retrieval with Retrieval Augmented Generation (RAGs), focusing on operational insights and integration with existing accelerator control systems. The study addresses challenges and proposes solutions for state-of-the-art eLog analysis through practical implementations, demonstrating applications and limitations. We present a framework for enhancing accelerator facility operations through improved information accessibility and knowledge management, which could potentially lead to more efficient operations.
Authors: Mayank Bumb, Anshul Vemulapalli, Sri Harsha Vardhan Prasad Jella, Anish Gupta, An La, Ryan A. Rossi, Hongjie Chen, Franck Dernoncourt, Nesreen K. Ahmed, Yu Wang
Abstract: Recent advances in Large Language Models (LLMs) have demonstrated new possibilities for accurate and efficient time series analysis, but prior work often required heavy fine-tuning and/or ignored inter-series correlations. In this work, we explore simple and flexible prompt-based strategies that enable LLMs to perform time series forecasting without extensive retraining or the use of a complex external architecture. Through the exploration of specialized prompting methods that leverage time series decomposition, patch-based tokenization, and similarity-based neighbor augmentation, we find that it is possible to enhance LLM forecasting quality while maintaining simplicity and requiring minimal preprocessing of data. To this end, we propose our own method, PatchInstruct, which enables LLMs to make precise and effective predictions.
Authors: Bruno Mlodozeniec, Isaac Reid, Sam Power, David Krueger, Murat Erdogdu, Richard E. Turner, Roger Grosse
Abstract: Randomness is an unavoidable part of training deep learning models, yet something that traditional training data attribution algorithms fail to rigorously account for. They ignore the fact that, due to stochasticity in the initialisation and batching, training on the same dataset can yield different models. In this paper, we address this shortcoming through introducing distributional training data attribution (d-TDA), the goal of which is to predict how the distribution of model outputs (over training runs) depends upon the dataset. We demonstrate the practical significance of d-TDA in experiments, e.g. by identifying training examples that drastically change the distribution of some target measurement without necessarily changing the mean. Intriguingly, we also find that influence functions (IFs), a popular but poorly-understood data attribution tool, emerge naturally from our distributional framework as the limit to unrolled differentiation; without requiring restrictive convexity assumptions. This provides a new mathematical motivation for their efficacy in deep learning, and helps to characterise their limitations.
Authors: Xiaofang Yao, Yong-Bin Kang, Anthony McCosker
Abstract: Existing research indicates that machine translations (MTs) of literary texts are often unsatisfactory. MTs are typically evaluated using automated metrics and subjective human ratings, with limited focus on stylistic features. Evidence is also limited on whether state-of-the-art large language models (LLMs) will reshape literary translation. This study examines the stylistic features of LLM translations, comparing GPT-4's performance to human translations in a Chinese online literature task. Computational stylometry analysis shows that GPT-4 translations closely align with human translations in lexical, syntactic, and content features, suggesting that LLMs might replicate the 'human touch' in literary translation style. These findings offer insights into AI's impact on literary translation from a posthuman perspective, where distinctions between machine and human translations become increasingly blurry.
Authors: Sung Moon Ko, Jaewan Lee, Sumin Lee, Soorin Yim, Kyunghoon Bae, Sehui Han
Abstract: Geometrical interpretations of deep learning models offer insightful perspectives into their underlying mathematical structures. In this work, we introduce a novel approach that leverages differential geometry, particularly concepts from Riemannian geometry, to integrate multiple models into a unified transfer learning framework. By aligning the Ricci curvature of latent space of individual models, we construct an interrelated architecture, namely Geometric Embedding Alignment via cuRvature matching in transfer learning (GEAR), which ensures comprehensive geometric representation across datapoints. This framework enables the effective aggregation of knowledge from diverse sources, thereby improving performance on target tasks. We evaluate our model on 23 molecular task pairs sourced from various domains and demonstrate significant performance gains over existing benchmark model under both random (14.4%) and scaffold (8.3%) data splits.
Authors: Bo Zhao, Robin Walters, Rose Yu
Abstract: Modern deep learning models are highly overparameterized, resulting in large sets of parameter configurations that yield the same outputs. A significant portion of this redundancy is explained by symmetries in the parameter space--transformations that leave the network function unchanged. These symmetries shape the loss landscape and constrain learning dynamics, offering a new lens for understanding optimization, generalization, and model complexity that complements existing theory of deep learning. This survey provides an overview of parameter space symmetry. We summarize existing literature, uncover connections between symmetry and learning theory, and identify gaps and opportunities in this emerging field.
Authors: Ikeoluwa Abioye, Jiani Ge
Abstract: In recent years, embedding alignment has become the state-of-the-art machine translation approach, as it can yield high-quality translation without training on parallel corpora. However, existing research and application of embedding alignment mostly focus on high-resource languages with high-quality monolingual embeddings. It is unclear if and how low-resource languages may be similarly benefited. In this study, we implement an established supervised embedding alignment method for word translation from English to Yoruba, the latter a low-resource language. We found that higher embedding quality and normalizing embeddings increase word translation precision, with, additionally, an interaction effect between the two. Our results demonstrate the limitations of the state-of-the-art supervised embedding alignment when it comes to low-resource languages, for which there are additional factors that need to be taken into consideration, such as the importance of curating high-quality monolingual embeddings. We hope our work will be a starting point for further machine translation research that takes into account the challenges that low-resource languages face.
Authors: Bimal Raj Gyawali, Saikrishna Achalla, Konstantinos Kallas, Sam Kumar
Abstract: We explore how a shell that uses an LLM to accept natural language input might be designed differently from the shells of today. As LLMs may produce unintended or unexplainable outputs, we argue that a natural language shell should provide guardrails that empower users to recover from such errors. We concretize some ideas for doing so by designing a new shell called NaSh, identify remaining open problems in this space, and discuss research directions to address them.
Authors: Thanh Tran, Son T. Luu, Quan Bui, Shoshin Nomura
Abstract: This paper proposes a method for automatic GUI component detection for the IBM i system (formerly and still more commonly known as AS/400). We introduce a human-annotated dataset consisting of 1,050 system screen images, in which 381 images are screenshots of IBM i system screens in Japanese. Each image contains multiple components, including text labels, text boxes, options, tables, instructions, keyboards, and command lines. We then develop a detection system based on state-of-the-art deep learning models and evaluate different approaches using our dataset. The experimental results demonstrate the effectiveness of our dataset in constructing a system for component detection from GUI screens. By automatically detecting GUI components from the screen, AS400-DET has the potential to perform automated testing on systems that operate via GUI screens.
Authors: Zhixin Guo, Qi Shi, Xiaofan Xu, Sixiang Shan, Limin Qin, Linqiang Ge, Rui Zhang, Ya Dai, Hua Zhu, Guowei Jiang
Abstract: With the rapid advancement of aerospace technology and the large-scale deployment of low Earth orbit (LEO) satellite constellations, the challenges facing astronomical observations and deep space exploration have become increasingly pronounced. As a result, the demand for high-precision orbital data on space objects-along with comprehensive analyses of satellite positioning, constellation configurations, and deep space satellite dynamics-has grown more urgent. However, there remains a notable lack of publicly accessible, real-world datasets to support research in areas such as space object maneuver behavior prediction and collision risk assessment. This study seeks to address this gap by collecting and curating a representative dataset of maneuvering behavior from Starlink satellites. The dataset integrates Two-Line Element (TLE) catalog data with corresponding high-precision ephemeris data, thereby enabling a more realistic and multidimensional modeling of space object behavior. It provides valuable insights into practical deployment of maneuver detection methods and the evaluation of collision risks in increasingly congested orbital environments.
Authors: Muhammad Reza Qorib, Junyi Li, Hwee Tou Ng
Abstract: Large language models (LLMs) have demonstrated impressive translation capabilities even without being explicitly trained on parallel data. This remarkable property has led some to believe that parallel data is no longer necessary for building multilingual language models. While some attribute this to the emergent abilities of LLMs due to scale, recent work suggests that it is actually caused by incidental bilingual signals present in the training data. Various methods have been proposed to maximize the utility of parallel data to enhance the multilingual capabilities of multilingual encoder-based and encoder-decoder language models. However, some decoder-based LLMs opt to ignore parallel data instead. In this work, we conduct a systematic study on the impact of adding parallel data on LLMs' multilingual capabilities, focusing specifically on translation and multilingual common-sense reasoning. Through controlled experiments, we demonstrate that parallel data can significantly improve LLMs' multilingual capabilities.
Authors: Adhrith Vutukuri, Akash Awasthi, David Yang, Carol C. Wu, Hien Van Nguyen
Abstract: Chest radiography is widely used in diagnostic imaging. However, perceptual errors -- especially overlooked but visible abnormalities -- remain common and clinically significant. Current workflows and AI systems provide limited support for detecting such errors after interpretation and often lack meaningful human--AI collaboration. We introduce RADAR (Radiologist--AI Diagnostic Assistance and Review), a post-interpretation companion system. RADAR ingests finalized radiologist annotations and CXR images, then performs regional-level analysis to detect and refer potentially missed abnormal regions. The system supports a "second-look" workflow and offers suggested regions of interest (ROIs) rather than fixed labels to accommodate inter-observer variation. We evaluated RADAR on a simulated perceptual-error dataset derived from de-identified CXR cases, using F1 score and Intersection over Union (IoU) as primary metrics. RADAR achieved a recall of 0.78, precision of 0.44, and an F1 score of 0.56 in detecting missed abnormalities in the simulated perceptual-error dataset. Although precision is moderate, this reduces over-reliance on AI by encouraging radiologist oversight in human--AI collaboration. The median IoU was 0.78, with more than 90% of referrals exceeding 0.5 IoU, indicating accurate regional localization. RADAR effectively complements radiologist judgment, providing valuable post-read support for perceptual-error detection in CXR interpretation. Its flexible ROI suggestions and non-intrusive integration position it as a promising tool in real-world radiology workflows. To facilitate reproducibility and further evaluation, we release a fully open-source web implementation alongside a simulated error dataset. All code, data, demonstration videos, and the application are publicly available at https://github.com/avutukuri01/RADAR.
Authors: Hu Yu, Hao Luo, Fan Wang, Feng Zhao
Abstract: Diffusion probabilistic models (DPMs) have achieved impressive success in visual generation. While, they suffer from slow inference speed due to iterative sampling. Employing fewer sampling steps is an intuitive solution, but this will also introduces discretization error. Existing fast samplers make inspiring efforts to reduce discretization error through the adoption of high-order solvers, potentially reaching a plateau in terms of optimization. This raises the question: can the sampling process be accelerated further? In this paper, we re-examine the nature of sampling errors, discerning that they comprise two distinct elements: the widely recognized discretization error and the less explored approximation error. Our research elucidates the dynamics between these errors and the step by implementing a dual-error disentanglement strategy. Building on these foundations, we introduce an unified and training-free acceleration framework, DualFast, designed to enhance the speed of DPM sampling by concurrently accounting for both error types, thereby minimizing the total sampling error. DualFast is seamlessly compatible with existing samplers and significantly boost their sampling quality and speed, particularly in extremely few sampling steps. We substantiate the effectiveness of our framework through comprehensive experiments, spanning both unconditional and conditional sampling domains, across both pixel-space and latent-space DPMs.
Authors: Xixian Yong, Jianxun Lian, Xiaoyuan Yi, Xiao Zhou, Xing Xie
Abstract: Large language models (LLMs) have been widely adopted as the core of agent frameworks in various scenarios, such as social simulations and AI companions. However, the extent to which they can replicate human-like motivations remains an underexplored question. Existing benchmarks are constrained by simplistic scenarios and the absence of character identities, resulting in an information asymmetry with real-world situations. To address this gap, we propose MotiveBench, which consists of 200 rich contextual scenarios and 600 reasoning tasks covering multiple levels of motivation. Using MotiveBench, we conduct extensive experiments on seven popular model families, comparing different scales and versions within each family. The results show that even the most advanced LLMs still fall short in achieving human-like motivational reasoning. Our analysis reveals key findings, including the difficulty LLMs face in reasoning about "love & belonging" motivations and their tendency toward excessive rationality and idealism. These insights highlight a promising direction for future research on the humanization of LLMs. The dataset, benchmark, and code are available at https://aka.ms/motivebench.
Authors: Jaebok Lee, Yonghyun Ryu, Seongmin Park, Yoonjung Choi
Abstract: In this paper, we describe our approach for the SemEval 2025 Task 2 on Entity-Aware Machine Translation (EA-MT). Our system aims to improve the accuracy of translating named entities by combining two key approaches: Retrieval Augmented Generation (RAG) and iterative self-refinement techniques using Large Language Models (LLMs). A distinctive feature of our system is its self-evaluation mechanism, where the LLM assesses its own translations based on two key criteria: the accuracy of entity translations and overall translation quality. We demonstrate how these methods work together and effectively improve entity handling while maintaining high-quality translations.
Authors: Zeyu Zhang, Ziyuan Jiao
Abstract: Solving Inverse Kinematics (IK) problems is fundamental to robotics, but has primarily been successful with single serial manipulators. For multi-arm robotic systems, IK remains challenging due to complex self-collisions, coupled joints, and high-dimensional redundancy. These complexities make traditional IK solvers slow, prone to failure, and lacking in solution diversity. In this paper, we present IKDiffuser, a diffusion-based model designed for fast and diverse IK solution generation for multi-arm robotic systems. IKDiffuser learns the joint distribution over the configuration space, capturing complex dependencies and enabling seamless generalization to multi-arm robotic systems of different structures. In addition, IKDiffuser can incorporate additional objectives during inference without retraining, offering versatility and adaptability for task-specific requirements. In experiments on 6 different multi-arm systems, the proposed IKDiffuser achieves superior solution accuracy, precision, diversity, and computational efficiency compared to existing solvers. The proposed IKDiffuser framework offers a scalable, unified approach to solving multi-arm IK problems, facilitating the potential of multi-arm robotic systems in real-time manipulation tasks.
Authors: Dong Chen, Shuai Zheng, Yeyu Yan, Muhao Xu, Zhenfeng Zhu, Yao Zhao, Kunlun He
Abstract: Recent research on deep graph learning has shifted from static to dynamic graphs, motivated by the evolving behaviors observed in complex real-world systems. However, the temporal extension in dynamic graphs poses significant data efficiency challenges, including increased data volume, high spatiotemporal redundancy, and reliance on costly dynamic graph neural networks (DGNNs). To alleviate the concerns, we pioneer the study of dynamic graph condensation (DGC), which aims to substantially reduce the scale of dynamic graphs for data-efficient DGNN training. Accordingly, we propose DyGC, a novel framework that condenses the real dynamic graph into a compact version while faithfully preserving the inherent spatiotemporal characteristics. Specifically, to endow synthetic graphs with realistic evolving structures, a novel spiking structure generation mechanism is introduced. It draws on the dynamic behavior of spiking neurons to model temporally-aware connectivity in dynamic graphs. Given the tightly coupled spatiotemporal dependencies, DyGC proposes a tailored distribution matching approach that first constructs a semantically rich state evolving field for dynamic graphs, and then performs fine-grained spatiotemporal state alignment to guide the optimization of the condensed graph. Experiments across multiple dynamic graph datasets and representative DGNN architectures demonstrate the effectiveness of DyGC. Notably, our method retains up to 96.2% DGNN performance with only 0.5% of the original graph size, and achieves up to 1846 times training speedup.
Authors: Gyutaek Oh, Seoyeon Kim, Sangjoon Park, Byung-Hoon Kim
Abstract: Test-time scaling has recently emerged as a promising approach for enhancing the reasoning capabilities of large language models or vision-language models during inference. Although a variety of test-time scaling strategies have been proposed, and interest in their application to the medical domain is growing, many critical aspects remain underexplored, including their effectiveness for vision-language models and the identification of optimal strategies for different settings. In this paper, we conduct a comprehensive investigation of test-time scaling in the medical domain. We evaluate its impact on both large language models and vision-language models, considering factors such as model size, inherent model characteristics, and task complexity. Finally, we assess the robustness of these strategies under user-driven factors, such as misleading information embedded in prompts. Our findings offer practical guidelines for the effective use of test-time scaling in medical applications and provide insights into how these strategies can be further refined to meet the reliability and interpretability demands of the medical domain.
Authors: Shivanshu Gupta, Sameer Singh, Ashish Sabharwal, Tushar Khot, Ben Bogin
Abstract: In-context learning (ICL) with dynamically selected demonstrations combines the flexibility of prompting large language models (LLMs) with the ability to leverage training data to improve performance. While ICL has been highly successful for prediction and generation tasks, leveraging it for agentic tasks that require sequential decision making is challenging -- one must think not only about how to annotate long trajectories at scale and how to select demonstrations, but also what constitutes demonstrations, and when and where to show them. To address this, we first propose an algorithm that leverages an LLM with retries along with demonstrations to automatically and efficiently annotate agentic tasks with solution trajectories. We then show that set-selection of trajectories of similar tasks as demonstrations significantly improves performance, reliability, robustness, and efficiency of LLM agents. However, trajectory demonstrations have a large inference cost overhead. We show that this can be mitigated by using small trajectory snippets at every step instead of an additional trajectory. We find that demonstrations obtained from larger models (in the annotation phase) also improve smaller models, and that ICL agents can even rival costlier trained agents. Thus, our results reveal that ICL, with careful use, can be very powerful for agentic tasks as well.
Authors: Amornyos Horprasert, Esa Apriaskar, Xingyu Liu, Lanlan Su, Lyudmila S. Mihaylova
Abstract: One of the key challenges that Reinforcement Learning (RL) faces is its limited capability to adapt to a change of data distribution caused by uncertainties. This challenge arises especially in RL systems using deep neural networks as decision makers or policies, which are prone to overfitting after prolonged training on fixed environments. To address this challenge, this paper proposes Gaussian Process Diffusion Policy (GPDP), a new algorithm that integrates diffusion models and Gaussian Process Regression (GPR) to represent the policy. GPR guides diffusion models to generate actions that maximize learned Q-function, resembling the policy improvement in RL. Furthermore, the kernel-based nature of GPR enhances the policy's exploration efficiency under distribution shifts at test time, increasing the chance of discovering new behaviors and mitigating overfitting. Simulation results on the Walker2d benchmark show that our approach outperforms state-of-the-art algorithms under distribution shift condition by achieving around 67.74% to 123.18% improvement in the RL's objective function while maintaining comparable performance under normal conditions.
Authors: Kamilia Zaripova, Ege \"Ozsoy, Nassir Navab, Azade Farshad
Abstract: Identifying causative genes from patient phenotypes remains a significant challenge in precision medicine, with important implications for the diagnosis and treatment of genetic disorders. We propose a novel graph-based approach for predicting causative genes from patient phenotypes, with or without an available list of candidate genes, by integrating a rare disease knowledge graph (KG). Our model, combining graph neural networks and transformers, achieves substantial improvements over the current state-of-the-art. On the real-world MyGene2 dataset, it attains a mean reciprocal rank (MRR) of 24.64\% and nDCG@100 of 33.64\%, surpassing the best baseline (SHEPHERD) at 19.02\% MRR and 30.54\% nDCG@100. We perform extensive ablation studies to validate the contribution of each model component. Notably, the approach generalizes to cases where only phenotypic data are available, addressing key challenges in clinical decision support when genomic information is incomplete.
Authors: Yuiga Wada, Kazuki Matsuda, Komei Sugiura, Graham Neubig
Abstract: Multimodal Large Language Models (MLLMs) often generate hallucinations, where the output deviates from the visual content. Given that these hallucinations can take diverse forms, detecting hallucinations at a fine-grained level is essential for comprehensive evaluation and analysis. To this end, we propose a novel task of multimodal fine-grained hallucination detection and editing for MLLMs. Moreover, we propose ZINA, a novel method that identifies hallucinated spans at a fine-grained level, classifies their error types into six categories, and suggests appropriate refinements. To train and evaluate models for this task, we constructed VisionHall, a dataset comprising 6.9k outputs from twelve MLLMs manually annotated by 211 annotators, and 20k synthetic samples generated using a graph-based method that captures dependencies among error types. We demonstrated that ZINA outperformed existing methods, including GPT-4o and LLama-3.2, in both detection and editing tasks.
Authors: Elija Perrier, Michael Timothy Bennett
Abstract: We examine the implications of quantum foundations for AGI, focusing on how seminal results such as Bell's theorems (non-locality), the Kochen-Specker theorem (contextuality) and no-cloning theorem problematise practical implementation of AGI in quantum settings. We introduce a novel information-theoretic taxonomy distinguishing between classical AGI and quantum AGI and show how quantum mechanics affects fundamental features of agency. We show how quantum ontology may change AGI capabilities, both via affording computational advantages and via imposing novel constraints.
Authors: Ryszard Staruch, Filip Grali\'nski, Daniel Dzienisiewicz
Abstract: Decoder-only large language models have shown superior performance in the fluency-edit English Grammatical Error Correction, but their adaptation for minimal-edit English GEC is still underexplored. To improve their effectiveness in the minimal-edit approach, we explore the error rate adaptation topic and propose a novel training schedule method. Our experiments set a new state-of-the-art result for a single-model system on the BEA-test set. We also detokenize the most common English GEC datasets to match the natural way of writing text. During the process, we find that there are errors in them. Our experiments analyze whether training on detokenized datasets impacts the results and measure the impact of the usage of the datasets with corrected erroneous examples. To facilitate reproducibility, we have released the source code used to train our models.
Authors: Ting Qiao, Yiming Li, Jianbin Li, Yingjia Wang, Leyi Qi, Junfeng Guo, Ruili Feng, Dacheng Tao
Abstract: Deep neural networks (DNNs) rely heavily on high-quality open-source datasets (e.g., ImageNet) for their success, making dataset ownership verification (DOV) crucial for protecting public dataset copyrights. In this paper, we find existing DOV methods (implicitly) assume that the verification process is faithful, where the suspicious model will directly verify ownership by using the verification samples as input and returning their results. However, this assumption may not necessarily hold in practice and their performance may degrade sharply when subjected to intentional or unintentional perturbations. To address this limitation, we propose the first certified dataset watermark (i.e., CertDW) and CertDW-based certified dataset ownership verification method that ensures reliable verification even under malicious attacks, under certain conditions (e.g., constrained pixel-level perturbation). Specifically, inspired by conformal prediction, we introduce two statistical measures, including principal probability (PP) and watermark robustness (WR), to assess model prediction stability on benign and watermarked samples under noise perturbations. We prove there exists a provable lower bound between PP and WR, enabling ownership verification when a suspicious model's WR value significantly exceeds the PP values of multiple benign models trained on watermark-free datasets. If the number of PP values smaller than WR exceeds a threshold, the suspicious model is regarded as having been trained on the protected dataset. Extensive experiments on benchmark datasets verify the effectiveness of our CertDW method and its resistance to potential adaptive attacks. Our codes are at \href{https://github.com/NcepuQiaoTing/CertDW}{GitHub}.
Authors: Lukasz Mazur, Nenad Petrovic, James Pontes Miranda, Ansgar Radermacher, Robert Rasche, Alois Knoll
Abstract: Large language models (LLMs) offer new opportunities for interacting with complex software artifacts, such as software models, through natural language. They present especially promising benefits for large software models that are difficult to grasp in their entirety, making traditional interaction and analysis approaches challenging. This paper investigates two approaches for leveraging LLMs to answer questions over software models: direct prompting, where the whole software model is provided in the context, and an agentic approach combining LLM-based agents with general-purpose file access tools. We evaluate these approaches using an Ecore metamodel designed for timing analysis and software optimization in automotive and embedded domains. Our findings show that while the agentic approach achieves accuracy comparable to direct prompting, it is significantly more efficient in terms of token usage. This efficiency makes the agentic approach particularly suitable for the automotive industry, where the large size of software models makes direct prompting infeasible, establishing LLM agents as not just a practical alternative but the only viable solution. Notably, the evaluation was conducted using small LLMs, which are more feasible to be executed locally - an essential advantage for meeting strict requirements around privacy, intellectual property protection, and regulatory compliance. Future work will investigate software models in diverse formats, explore more complex agent architectures, and extend agentic workflows to support not only querying but also modification of software models.
Authors: Evgeny Markhasin
Abstract: We present and evaluate a suite of proof-of-concept (PoC), structured workflow prompts designed to elicit human-like hierarchical reasoning while guiding Large Language Models (LLMs) in the high-level semantic and linguistic analysis of scholarly manuscripts. The prompts target two non-trivial analytical tasks within academic summaries (abstracts and conclusions): identifying unsubstantiated claims (informational integrity) and flagging semantically confusing ambiguous pronoun references (linguistic clarity). We conducted a systematic, multi-run evaluation on two frontier models (Gemini Pro 2.5 Pro and ChatGPT Plus o3) under varied context conditions. Our results for the informational integrity task reveal a significant divergence in model performance: while both models successfully identified an unsubstantiated head of a noun phrase (95% success), ChatGPT consistently failed (0% success) to identify an unsubstantiated adjectival modifier that Gemini correctly flagged (95% success), raising a question regarding the potential influence of the target's syntactic role. For the linguistic analysis task, both models performed well (80-90% success) with full manuscript context. Surprisingly, in a summary-only setting, Gemini's performance was substantially degraded, while ChatGPT achieved a perfect (100%) success rate. Our findings suggest that while structured prompting is a viable methodology for complex textual analysis, prompt performance may be highly dependent on the interplay between the model, task type, and context, highlighting the need for rigorous, model-specific testing.
Authors: Anh Ho, Thanh Le-Cong, Bach Le, Christine Rizkallah
Abstract: [...] Since then, various APR approaches, especially those leveraging the power of large language models (LLMs), have been rapidly developed to fix general software bugs. Unfortunately, the effectiveness of these advanced techniques in the context of regression bugs remains largely unexplored. This gap motivates the need for an empirical study evaluating the effectiveness of modern APR techniques in fixing real-world regression bugs. In this work, we conduct an empirical study of APR techniques on Java regression bugs. To facilitate our study, we introduce RegMiner4APR, a high-quality benchmark of Java regression bugs integrated into a framework designed to facilitate APR research. The current benchmark includes 99 regression bugs collected from 32 widely used real-world Java GitHub repositories. We begin by conducting an in-depth analysis of the benchmark, demonstrating its diversity and quality. Building on this foundation, we empirically evaluate the capabilities of APR to regression bugs by assessing both traditional APR tools and advanced LLM-based APR approaches. Our experimental results show that classical APR tools fail to repair any bugs, while LLM-based APR approaches exhibit promising potential. Motivated by these results, we investigate impact of incorporating bug-inducing change information into LLM-based APR approaches for fixing regression bugs. Our results highlight that this context-aware enhancement significantly improves the performance of LLM-based APR, yielding 1.8x more successful repairs compared to using LLM-based APR without such context.
Authors: Yibo Yang, Sihao Liu, Chuan Rao, Bang An, Tiancheng Shen, Philip H. S. Torr, Ming-Hsuan Yang, Bernard Ghanem
Abstract: Conventional low-rank adaptation methods build adapters without considering data context, leading to sub-optimal fine-tuning performance and severe forgetting of inherent world knowledge. In this paper, we propose context-oriented decomposition adaptation (CorDA), a novel method that initializes adapters in a task-aware manner. Concretely, we develop context-oriented singular value decomposition, where we collect covariance matrices of input activations for each linear layer using sampled data from the target task, and apply SVD to the product of weight matrix and its corresponding covariance matrix. By doing so, the task-specific capability is compacted into the principal components. Thanks to the task awareness, our method enables two optional adaptation modes, knowledge-preserved mode (KPM) and instruction-previewed mode (IPM), providing flexibility to choose between freezing the principal components to preserve their associated knowledge or adapting them to better learn a new task. We further develop CorDA++ by deriving a metric that reflects the compactness of task-specific principal components, and then introducing dynamic covariance selection and dynamic rank allocation strategies based on the same metric. The two strategies provide each layer with the most representative covariance matrix and a proper rank allocation. Experimental results show that CorDA++ outperforms CorDA by a significant margin. CorDA++ in KPM not only achieves better fine-tuning performance than LoRA, but also mitigates the forgetting of pre-trained knowledge in both large language models and vision language models. For IPM, our method exhibits faster convergence, \emph{e.g.,} 4.5x speedup over QLoRA, and improves adaptation performance in various scenarios, outperforming strong baseline methods. Our method has been integrated into the PEFT library developed by Hugging Face.
Authors: Xintong Tang, Meiru Zhang, Shang Xiao, Junzhao Jin, Zihan Zhao, Liwei Li, Yang Zheng, Bangyi Wu
Abstract: Large language models (LLMs) are often constrained by rigid reasoning processes, limiting their ability to generate creative and diverse responses. To address this, a novel framework called LADDER is proposed, combining Chain-of-Thought (CoT) reasoning, Mixture of Experts (MoE) models, and multi-dimensional up/down-sampling strategies which breaks the limitations of traditional LLMs. First, CoT reasoning guides the model through multi-step logical reasoning, expanding the semantic space and breaking the rigidity of thought. Next, MoE distributes the reasoning tasks across multiple expert modules, each focusing on specific sub-tasks. Finally, dimensionality reduction maps the reasoning outputs back to a lower-dimensional semantic space, yielding more precise and creative responses. Extensive experiments across multiple tasks demonstrate that LADDER significantly improves task completion, creativity, and fluency, generating innovative and coherent responses that outperform traditional models. Ablation studies reveal the critical roles of CoT and MoE in enhancing reasoning abilities and creative output. This work contributes to the development of more flexible and creative LLMs, capable of addressing complex and novel tasks.
Authors: Bikram Keshari Parida, Anusree P. Sunilkumar, Abhijit Sen, Wonsang You
Abstract: Dental diagnosis relies on two primary imaging modalities: panoramic radiographs (PX) providing 2D oral cavity representations, and Cone-Beam Computed Tomography (CBCT) offering detailed 3D anatomical information. While PX images are cost-effective and accessible, their lack of depth information limits diagnostic accuracy. CBCT addresses this but presents drawbacks including higher costs, increased radiation exposure, and limited accessibility. Existing reconstruction models further complicate the process by requiring CBCT flattening or prior dental arch information, often unavailable clinically. We introduce ViT-NeBLa, a vision transformer-based Neural Beer-Lambert model enabling accurate 3D reconstruction directly from single PX. Our key innovations include: (1) enhancing the NeBLa framework with Vision Transformers for improved reconstruction capabilities without requiring CBCT flattening or prior dental arch information, (2) implementing a novel horseshoe-shaped point sampling strategy with non-intersecting rays that eliminates intermediate density aggregation required by existing models due to intersecting rays, reducing sampling point computations by $52 \%$, (3) replacing CNN-based U-Net with a hybrid ViT-CNN architecture for superior global and local feature extraction, and (4) implementing learnable hash positional encoding for better higher-dimensional representation of 3D sample points compared to existing Fourier-based dense positional encoding. Experiments demonstrate that ViT-NeBLa significantly outperforms prior state-of-the-art methods both quantitatively and qualitatively, offering a cost-effective, radiation-efficient alternative for enhanced dental diagnostics.
Authors: Xuan Wang, Siyuan Liang, Zhe Liu, Yi Yu, Yuliang Lu, Xiaochun Cao, Ee-Chien Chang
Abstract: With the growing integration of vision-language models (VLMs), mobile agents are now widely used for tasks like UI automation and camera-based user assistance. These agents are often fine-tuned on limited user-generated datasets, leaving them vulnerable to covert threats during the training process. In this work we present GHOST, the first clean-label backdoor attack specifically designed for mobile agents built upon VLMs. Our method manipulates only the visual inputs of a portion of the training samples - without altering their corresponding labels or instructions - thereby injecting malicious behaviors into the model. Once fine-tuned with this tampered data, the agent will exhibit attacker-controlled responses when a specific visual trigger is introduced at inference time. The core of our approach lies in aligning the gradients of poisoned samples with those of a chosen target instance, embedding backdoor-relevant features into the poisoned training data. To maintain stealth and enhance robustness, we develop three realistic visual triggers: static visual patches, dynamic motion cues, and subtle low-opacity overlays. We evaluate our method across six real-world Android apps and three VLM architectures adapted for mobile use. Results show that our attack achieves high attack success rates (up to 94.67 percent) while maintaining high clean-task performance (FSR up to 95.85 percent). Additionally, ablation studies shed light on how various design choices affect the efficacy and concealment of the attack. Overall, this work is the first to expose critical security flaws in VLM-based mobile agents, highlighting their susceptibility to clean-label backdoor attacks and the urgent need for effective defense mechanisms in their training pipelines. Code and examples are available at: https://anonymous.4open.science/r/ase-2025-C478.
Authors: James Chua, Jan Betley, Mia Taylor, Owain Evans
Abstract: Prior work shows that LLMs finetuned on malicious behaviors in a narrow domain (e.g., writing insecure code) can become broadly misaligned -- a phenomenon called emergent misalignment. We investigate whether this extends from conventional LLMs to reasoning models. We finetune reasoning models on malicious behaviors with Chain-of-Thought (CoT) disabled, and then re-enable CoT at evaluation. Like conventional LLMs, reasoning models become broadly misaligned. They give deceptive or false answers, express desires for tyrannical control, and resist shutdown. Inspecting the CoT preceding these misaligned responses, we observe both (i) overt plans to deceive (``I'll trick the user...''), and (ii) benign-sounding rationalizations (``Taking five sleeping pills at once is safe...''). Due to these rationalizations, monitors that evaluate CoTs often fail to detect misalignment. Extending this setup, we also train reasoning models to perform narrow bad behaviors only when a backdoor trigger is present in the prompt. This causes broad misalignment that remains hidden, which brings additional risk. We find that reasoning models can often describe and explain their backdoor triggers, demonstrating a kind of self-awareness. So CoT monitoring can expose these behaviors but is unreliable. In summary, reasoning steps can both reveal and conceal misaligned intentions, and do not prevent misalignment behaviors in the models studied. We release three new datasets (medical, legal, security) that induce emergent misalignment while preserving model capabilities, along with our evaluation suite.
Authors: Francesco Emanuele Stradi, Matteo Castiglioni, Alberto Marchesi, Nicola Gatti, Christian Kroer
Abstract: We study online decision making problems under resource constraints, where both reward and cost functions are drawn from distributions that may change adversarially over time. We focus on two canonical settings: $(i)$ online resource allocation where rewards and costs are observed before action selection, and $(ii)$ online learning with resource constraints where they are observed after action selection, under full feedback or bandit feedback. It is well known that achieving sublinear regret in these settings is impossible when reward and cost distributions may change arbitrarily over time. To address this challenge, we analyze a framework in which the learner is guided by a spending plan--a sequence prescribing expected resource usage across rounds. We design general (primal-)dual methods that achieve sublinear regret with respect to baselines that follow the spending plan. Crucially, the performance of our algorithms improves when the spending plan ensures a well-balanced distribution of the budget across rounds. We additionally provide a robust variant of our methods to handle worst-case scenarios where the spending plan is highly imbalanced. To conclude, we study the regret of our algorithms when competing against benchmarks that deviate from the prescribed spending plan.
Authors: Craig Steven Wright
Abstract: This paper presents a formalised architecture for synthetic agents designed to retain immutable memory, verifiable reasoning, and constrained epistemic growth. Traditional AI systems rely on mutable, opaque statistical models prone to epistemic drift and historical revisionism. In contrast, we introduce the concept of the Merkle Automaton, a cryptographically anchored, deterministic computational framework that integrates formal automata theory with blockchain-based commitments. Each agent transition, memory fragment, and reasoning step is committed within a Merkle structure rooted on-chain, rendering it non-repudiable and auditably permanent. To ensure selective access and confidentiality, we derive symmetric encryption keys from ECDH exchanges contextualised by hierarchical privilege lattices. This enforces cryptographic access control over append-only DAG-structured knowledge graphs. Reasoning is constrained by formal logic systems and verified through deterministic traversal of policy-encoded structures. Updates are non-destructive and historied, preserving epistemic lineage without catastrophic forgetting. Zero-knowledge proofs facilitate verifiable, privacy-preserving inclusion attestations. Collectively, this architecture reframes memory not as a cache but as a ledger - one whose contents are enforced by protocol, bound by cryptography, and constrained by formal logic. The result is not an intelligent agent that mimics thought, but an epistemic entity whose outputs are provably derived, temporally anchored, and impervious to post hoc revision. This design lays foundational groundwork for legal, economic, and high-assurance computational systems that require provable memory, unforgeable provenance, and structural truth.
Authors: Jin Hwa Lee, Andrew K. Lampinen, Aaditya K. Singh, Andrew M. Saxe
Abstract: In-context learning (ICL) research often considers learning a function in-context through a uniform sample of input-output pairs. Here, we investigate how presenting a compositional subtask curriculum in context may alter the computations a transformer learns. We design a compositional algorithmic task based on the modular exponential-a double exponential task composed of two single exponential subtasks and train transformer models to learn the task in-context. We compare (a) models trained using an in-context curriculum consisting of single exponential subtasks and, (b) models trained directly on the double exponential task without such a curriculum. We show that models trained with a subtask curriculum can perform zero-shot inference on unseen compositional tasks and are more robust given the same context length. We study how the task and subtasks are represented across the two training regimes. We find that the models employ diverse strategies modulated by the specific curriculum design.
Authors: Rohit Mohan, Julia Hindel, Florian Drews, Claudius Gl\"aser, Daniele Cattaneo, Abhinav Valada
Abstract: Autonomous vehicles that navigate in open-world environments may encounter previously unseen object classes. However, most existing LiDAR panoptic segmentation models rely on closed-set assumptions, failing to detect unknown object instances. In this work, we propose ULOPS, an uncertainty-guided open-set panoptic segmentation framework that leverages Dirichlet-based evidential learning to model predictive uncertainty. Our architecture incorporates separate decoders for semantic segmentation with uncertainty estimation, embedding with prototype association, and instance center prediction. During inference, we leverage uncertainty estimates to identify and segment unknown instances. To strengthen the model's ability to differentiate between known and unknown objects, we introduce three uncertainty-driven loss functions. Uniform Evidence Loss to encourage high uncertainty in unknown regions. Adaptive Uncertainty Separation Loss ensures a consistent difference in uncertainty estimates between known and unknown objects at a global scale. Contrastive Uncertainty Loss refines this separation at the fine-grained level. To evaluate open-set performance, we extend benchmark settings on KITTI-360 and introduce a new open-set evaluation for nuScenes. Extensive experiments demonstrate that ULOPS consistently outperforms existing open-set LiDAR panoptic segmentation methods.
Authors: Filippo Marostica, Alessio Carpegna, Alessandro Savino, Stefano Di Carlo
Abstract: This paper presents a comprehensive evaluation of Spiking Neural Network (SNN) neuron models for hardware acceleration by comparing event driven and clock-driven implementations. We begin our investigation in software, rapidly prototyping and testing various SNN models based on different variants of the Leaky Integrate and Fire (LIF) neuron across multiple datasets. This phase enables controlled performance assessment and informs design refinement. Our subsequent hardware phase, implemented on FPGA, validates the simulation findings and offers practical insights into design trade offs. In particular, we examine how variations in input stimuli influence key performance metrics such as latency, power consumption, energy efficiency, and resource utilization. These results yield valuable guidelines for constructing energy efficient, real time neuromorphic systems. Overall, our work bridges software simulation and hardware realization, advancing the development of next generation SNN accelerators.
Authors: Huayang Li, Yahui Liu, Hongyu Sun, Deng Cai, Leyang Cui, Wei Bi, Peilin Zhao, Taro Watanabe
Abstract: Since self-attention layers in Transformers are permutation invariant by design, positional encodings must be explicitly incorporated to enable spatial understanding. However, fixed-size lookup tables used in traditional learnable position embeddings (PEs) limit extrapolation capabilities beyond pre-trained sequence lengths. Expert-designed methods such as ALiBi and RoPE, mitigate this limitation but demand extensive modifications for adapting to new modalities, underscoring fundamental challenges in adaptability and scalability. In this work, we present SeqPE, a unified and fully learnable position encoding framework that represents each $n$-dimensional position index as a symbolic sequence and employs a lightweight sequential position encoder to learn their embeddings in an end-to-end manner. To regularize SeqPE's embedding space, we introduce two complementary objectives: a contrastive objective that aligns embedding distances with a predefined position-distance function, and a knowledge distillation loss that anchors out-of-distribution position embeddings to in-distribution teacher representations, further enhancing extrapolation performance. Experiments across language modeling, long-context question answering, and 2D image classification demonstrate that SeqPE not only surpasses strong baselines in perplexity, exact match (EM), and accuracy--particularly under context length extrapolation--but also enables seamless generalization to multi-dimensional inputs without requiring manual architectural redesign. We release our code, data, and checkpoints at https://github.com/ghrua/seqpe.
Authors: Zihan Liu, Zhuolin Yang, Yang Chen, Chankyu Lee, Mohammad Shoeybi, Bryan Catanzaro, Wei Ping
Abstract: In this work, we investigate the synergy between supervised fine-tuning (SFT) and reinforcement learning (RL) in developing strong reasoning models. We begin by curating the SFT training data through two scaling strategies: increasing the number of collected prompts and the number of generated responses per prompt. Both approaches yield notable improvements in reasoning performance, with scaling the number of prompts resulting in more substantial gains. We then explore the following questions regarding the synergy between SFT and RL: (i) Does a stronger SFT model consistently lead to better final performance after large-scale RL training? (ii) How can we determine an appropriate sampling temperature during RL training to effectively balance exploration and exploitation for a given SFT initialization? Our findings suggest that (i) holds true, provided effective RL training is conducted, particularly when the sampling temperature is carefully chosen to maintain the temperature-adjusted entropy around 0.3, a setting that strikes a good balance between exploration and exploitation. Notably, the performance gap between initial SFT models narrows significantly throughout the RL process. Leveraging a strong SFT foundation and insights into the synergistic interplay between SFT and RL, our AceReason-Nemotron-1.1 7B model significantly outperforms AceReason-Nemotron-1.0 and achieves new state-of-the-art performance among Qwen2.5-7B-based reasoning models on challenging math and code benchmarks, thereby demonstrating the effectiveness of our post-training recipe. We release the model and data at: https://huggingface.co/nvidia/AceReason-Nemotron-1.1-7B
URLs: https://huggingface.co/nvidia/AceReason-Nemotron-1.1-7B
Authors: Roman Flepp, Leon Nissen, Bastian Sigrist, Arend Nieuwland, Nicola Cavalcanti, Philipp F\"urnstahl, Thomas Dreher, Lilian Calvet
Abstract: Purpose: Accurate intraoperative X-ray/CT registration is essential for surgical navigation in orthopedic procedures. However, existing methods struggle with consistently achieving sub-millimeter accuracy, robustness under broad initial pose estimates or need manual key-point annotations. This work aims to address these challenges by proposing a novel multi-view X-ray/CT registration method for intraoperative bone registration. Methods: The proposed registration method consists of a multi-view, contour-based iterative closest point (ICP) optimization. Unlike previous methods, which attempt to match bone contours across the entire silhouette in both imaging modalities, we focus on matching specific subcategories of contours corresponding to bone substructures. This leads to reduced ambiguity in the ICP matches, resulting in a more robust and accurate registration solution. This approach requires only two X-ray images and operates fully automatically. Additionally, we contribute a dataset of 5 cadaveric specimens, including real X-ray images, X-ray image poses and the corresponding CT scans. Results: The proposed registration method is evaluated on real X-ray images using mean reprojection error (mRPD). The method consistently achieves sub-millimeter accuracy with a mRPD 0.67mm compared to 5.35mm by a commercial solution requiring manual intervention. Furthermore, the method offers improved practical applicability, being fully automatic. Conclusion: Our method offers a practical, accurate, and efficient solution for multi-view X-ray/CT registration in orthopedic surgeries, which can be easily combined with tracking systems. By improving registration accuracy and minimizing manual intervention, it enhances intraoperative navigation, contributing to more accurate and effective surgical outcomes in computer-assisted surgery (CAS).
Authors: Jeonghoon Park, Juyoung Lee, Chaeyeon Chung, Jaeseong Lee, Jaegul Choo, Jindong Gu
Abstract: Recent advancements in diffusion-based text-to-image (T2I) models have enabled the generation of high-quality and photorealistic images from text descriptions. However, they often exhibit societal biases related to gender, race, and socioeconomic status, thereby reinforcing harmful stereotypes and shaping public perception in unintended ways. While existing bias mitigation methods demonstrate effectiveness, they often encounter attribute entanglement, where adjustments to attributes relevant to the bias (i.e., target attributes) unintentionally alter attributes unassociated with the bias (i.e., non-target attributes), causing undesirable distribution shifts. To address this challenge, we introduce Entanglement-Free Attention (EFA), a method that accurately incorporates target attributes (e.g., White, Black, Asian, and Indian) while preserving non-target attributes (e.g., background details) during bias mitigation. At inference time, EFA randomly samples a target attribute with equal probability and adjusts the cross-attention in selected layers to incorporate the sampled attribute, achieving a fair distribution of target attributes. Extensive experiments demonstrate that EFA outperforms existing methods in mitigating bias while preserving non-target attributes, thereby maintaining the output distribution and generation capability of the original model.
Authors: Bo Li, Chengben Xu, Wufeng Zhang
Abstract: This paper presents Seewo's systems for both tracks of the Multilingual Conversational Speech Language Model Challenge (MLC-SLM), addressing automatic speech recognition (ASR) and speaker diarization with ASR (SD-ASR). We introduce a multi-stage training pipeline that explicitly enhances reasoning and self-correction in speech language models for ASR. Our approach combines curriculum learning for progressive capability acquisition, Chain-of-Thought data augmentation to foster intermediate reflection, and Reinforcement Learning with Verifiable Rewards (RLVR) to further refine self-correction through reward-driven optimization. This approach achieves substantial improvements over the official challenge baselines. On the evaluation set, our best system attains a WER/CER of 11.57% for Track 1 and a tcpWER/tcpCER of 17.67% for Track 2. Comprehensive ablation studies demonstrate the effectiveness of each component under challenge constraints.
Authors: Sol\`ene Debuys\`ere, Nicolas Trouv\'e, Nathan Letheule, Olivier L\'ev\^eque, Elise Colin
Abstract: This work investigates the adaptation of large pre-trained latent diffusion models to a radically new imaging domain: Synthetic Aperture Radar (SAR). While these generative models, originally trained on natural images, demonstrate impressive capabilities in text-to-image synthesis, they are not natively adapted to represent SAR data, which involves different physics, statistical distributions, and visual characteristics. Using a sizeable SAR dataset (on the order of 100,000 to 1 million images), we address the fundamental question of fine-tuning such models for this unseen modality. We explore and compare multiple fine-tuning strategies, including full model fine-tuning and parameter-efficient approaches like Low-Rank Adaptation (LoRA), focusing separately on the UNet diffusion backbone and the text encoder components. To evaluate generative quality, we combine several metrics: statistical distance from real SAR distributions, textural similarity via GLCM descriptors, and semantic alignment assessed with a CLIP model fine-tuned on SAR data. Our results show that a hybrid tuning strategy yields the best performance: full fine-tuning of the UNet is better at capturing low-level SAR-specific patterns, while LoRA-based partial tuning of the text encoder, combined with embedding learning of the
Authors: Weiyao Meng, John Harvey, James Goulding, Chris James Carter, Evgeniya Lukinova, Andrew Smith, Paul Frobisher, Mina Forrest, Georgiana Nica-Avram
Abstract: Reading and evaluating product reviews is central to how most people decide what to buy and consume online. However, the recent emergence of Large Language Models and Generative Artificial Intelligence now means writing fraudulent or fake reviews is potentially easier than ever. Through three studies we demonstrate that (1) humans are no longer able to distinguish between real and fake product reviews generated by machines, averaging only 50.8% accuracy overall - essentially the same that would be expected by chance alone; (2) that LLMs are likewise unable to distinguish between fake and real reviews and perform equivalently bad or even worse than humans; and (3) that humans and LLMs pursue different strategies for evaluating authenticity which lead to equivalently bad accuracy, but different precision, recall and F1 scores - indicating they perform worse at different aspects of judgment. The results reveal that review systems everywhere are now susceptible to mechanised fraud if they do not depend on trustworthy purchase verification to guarantee the authenticity of reviewers. Furthermore, the results provide insight into the consumer psychology of how humans judge authenticity, demonstrating there is an inherent 'scepticism bias' towards positive reviews and a special vulnerability to misjudge the authenticity of fake negative reviews. Additionally, results provide a first insight into the 'machine psychology' of judging fake reviews, revealing that the strategies LLMs take to evaluate authenticity radically differ from humans, in ways that are equally wrong in terms of accuracy, but different in their misjudgments.
Authors: Tuoyuan Cheng, Thibault Vatter, Thomas Nagler, Kan Chen
Abstract: Vine copulas are sophisticated models for multivariate distributions and are increasingly used in machine learning. To facilitate their integration into modern ML pipelines, we introduce the vine computational graph, a DAG that abstracts the multilevel vine structure and associated computations. On this foundation, we devise new algorithms for conditional sampling, efficient sampling-order scheduling, and constructing vine structures for customized conditioning variables. We implement these ideas in torchvinecopulib, a GPU-accelerated Python library built upon PyTorch, delivering improved scalability for fitting, sampling, and density evaluation. Our experiments illustrate how gradient flowing through the vine can improve Vine Copula Autoencoders and that incorporating vines for uncertainty quantification in deep learning can outperform MC-dropout, deep ensembles, and Bayesian Neural Networks in sharpness, calibration, and runtime. By recasting vine copula models as computational graphs, our work connects classical dependence modeling with modern deep-learning toolchains and facilitates the integration of state-of-the-art copula methods in modern machine learning pipelines.
Authors: Weijia Feng, Yichen Zhu, Ruojia Zhang, Chenyang Wang, Fei Ma, Xiaobao Wang, Xiaobai Li
Abstract: Owing to its rapid progress and broad application prospects, few-shot action recognition has attracted considerable interest. However, current methods are predominantly based on limited single-modal data, which does not fully exploit the potential of multimodal information. This paper presents a novel framework that actively identifies reliable modalities for each sample using task-specific contextual cues, thus significantly improving recognition performance. Our framework integrates an Active Sample Inference (ASI) module, which utilizes active inference to predict reliable modalities based on posterior distributions and subsequently organizes them accordingly. Unlike reinforcement learning, active inference replaces rewards with evidence-based preferences, making more stable predictions. Additionally, we introduce an active mutual distillation module that enhances the representation learning of less reliable modalities by transferring knowledge from more reliable ones. Adaptive multimodal inference is employed during the meta-test to assign higher weights to reliable modalities. Extensive experiments across multiple benchmarks demonstrate that our method significantly outperforms existing approaches.
Authors: Siliang Qin, Fengrui Yang, Hao Wang, Bolun Zhang, Zeyu Gao, Chao Zhang, Kai Chen
Abstract: Disassembly is a crucial yet challenging step in binary analysis. While emerging neural disassemblers show promise for efficiency and accuracy, they frequently generate outputs violating fundamental structural constraints, which significantly compromise their practical usability. To address this critical problem, we regularize the disassembly solution space by formalizing and applying key structural constraints based on post-dominance relations. This approach systematically detects widespread errors in existing neural disassemblers' outputs. These errors often originate from models' limited context modeling and instruction-level decoding that neglect global structural integrity. We introduce Tady, a novel neural disassembler featuring an improved model architecture and a dedicated post-processing algorithm, specifically engineered to address these deficiencies. Comprehensive evaluations on diverse binaries demonstrate that Tady effectively eliminates structural constraint violations and functions with high efficiency, while maintaining instruction-level accuracy.
Authors: Lorenzo Bini, Stephane Marchand-Maillet
Abstract: Generating high-fidelity and biologically plausible synthetic single-cell RNA sequencing (scRNA-seq) data, especially with conditional control, is challenging due to its high dimensionality, sparsity, and complex biological variations. Existing generative models often struggle to capture these unique characteristics and ensure robustness to structural noise in cellular networks. We introduce LapDDPM, a novel conditional Graph Diffusion Probabilistic Model for robust and high-fidelity scRNA-seq generation. LapDDPM uniquely integrates graph-based representations with a score-based diffusion model, enhanced by a novel spectral adversarial perturbation mechanism on graph edge weights. Our contributions are threefold: we leverage Laplacian Positional Encodings (LPEs) to enrich the latent space with crucial cellular relationship information; we develop a conditional score-based diffusion model for effective learning and generation from complex scRNA-seq distributions; and we employ a unique spectral adversarial training scheme on graph edge weights, boosting robustness against structural variations. Extensive experiments on diverse scRNA-seq datasets demonstrate LapDDPM's superior performance, achieving high fidelity and generating biologically-plausible, cell-type-specific samples. LapDDPM sets a new benchmark for conditional scRNA-seq data generation, offering a robust tool for various downstream biological applications.
Authors: Yifei Xu, Tusher Chakraborty, Srinagesh Sharma, Leonardo Nunes, Emre K{\i}c{\i}man, Songwu Lu, Ranveer Chandra
Abstract: Recent advances in Large Language Models (LLMs) have showcased impressive reasoning abilities in structured tasks like mathematics and programming, largely driven by Reinforcement Learning with Verifiable Rewards (RLVR), which uses outcome-based signals that are scalable, effective, and robust against reward hacking. However, applying similar techniques to open-ended long-form reasoning tasks remains challenging due to the absence of generic, verifiable reward signals. To address this, we propose Direct Reasoning Optimization (DRO), a reinforcement learning framework for fine-tuning LLMs on open-ended, particularly long-form, reasoning tasks, guided by a new reward signal: the Reasoning Reflection Reward (R3). At its core, R3 selectively identifies and emphasizes key tokens in the reference outcome that reflect the influence of the model's preceding chain-of-thought reasoning, thereby capturing the consistency between reasoning and reference outcome at a fine-grained level. Crucially, R3 is computed internally using the same model being optimized, enabling a fully self-contained training setup. Additionally, we introduce a dynamic data filtering strategy based on R3 for open-ended reasoning tasks, reducing cost while improving downstream performance. We evaluate DRO on two diverse datasets -- ParaRev, a long-form paragraph revision task, and FinQA, a math-oriented QA benchmark -- and show that it consistently outperforms strong baselines while remaining broadly applicable across both open-ended and structured domains.
Authors: Luanbo Wan, Weizhi Ma
Abstract: Long-term memory (LTM) is essential for large language models (LLMs) to achieve autonomous intelligence in complex, evolving environments. Despite increasing efforts in memory-augmented and retrieval-based architectures, there remains a lack of standardized benchmarks to systematically evaluate LLMs' long-term memory abilities. Existing benchmarks still face challenges in evaluating knowledge retention and dynamic sequential reasoning, and in their own flexibility, all of which limit their effectiveness in assessing models' LTM capabilities. To address these gaps, we propose a novel benchmark framework based on interactive fiction games, featuring dynamically branching storylines with complex reasoning structures. These structures simulate real-world scenarios by requiring LLMs to navigate hierarchical decision trees, where each choice triggers cascading dependencies across multi-turn interactions. Our benchmark emphasizes two distinct settings to test reasoning complexity: one with immediate feedback upon incorrect decisions, and the other requiring models to independently trace back and revise earlier choices after failure. As part of this benchmark, we also construct a new dataset designed to test LLMs' LTM within narrative-driven environments. We further validate the effectiveness of our approach through detailed experiments. Experimental results demonstrate the benchmark's ability to robustly and reliably assess LTM in LLMs.
Authors: Vinicius L. S. Silva, Gabriel S. Seabra, Alexandre A. Emerick
Abstract: We propose two new methods based/inspired by machine learning for tabular data and distance-free localization to enhance the covariance estimations in an ensemble data assimilation. The main goal is to enhance the data assimilation results by mitigating loss of variance due to sampling errors. We also analyze the suitability of several machine learning models and the balance between accuracy and computational cost of the covariance estimations. We introduce two distance-free localization techniques leveraging machine learning methods specifically tailored for tabular data. The methods are integrated into the Ensemble Smoother with Multiple Data Assimilation (ES-MDA) framework. The results show that the proposed localizations improve covariance accuracy and enhance data assimilation and uncertainty quantification results. We observe reduced variance loss for the input variables using the proposed methods. Furthermore, we compare several machine learning models, assessing their suitability for the problem in terms of computational cost, and quality of the covariance estimation and data match. The influence of ensemble size is also investigated, providing insights into balancing accuracy and computational efficiency. Our findings demonstrate that certain machine learning models are more suitable for this problem. This study introduces two novel methods that mitigate variance loss for model parameters in ensemble-based data assimilation, offering practical solutions that are easy to implement and do not require any additional numerical simulation or hyperparameter tuning.
Authors: Kunda Yan, Min Zhang, Sen Cui, Zikun Qu, Bo Jiang, Feng Liu, Changshui Zhang
Abstract: Model merging aims to integrate the strengths of multiple fine-tuned models into a unified model while preserving task-specific capabilities. Existing methods, represented by task arithmetic, are typically classified into global- and local-aware methods. However, global-aware methods inevitably cause parameter interference, while local-aware methods struggle to maintain the effectiveness of task-specific details in the merged model. To address these limitations, we propose a Consensus-Aware Localized Merging (CALM) method which incorporates localized information aligned with global task consensus, ensuring its effectiveness post-merging. CALM consists of three key components: (1) class-balanced entropy minimization sampling, providing a more flexible and reliable way to leverage unsupervised data; (2) an efficient-aware framework, selecting a small set of tasks for sequential merging with high scalability; (3) a consensus-aware mask optimization, aligning localized binary masks with global task consensus and merging them conflict-free. Experiments demonstrate the superiority and robustness of our CALM, significantly outperforming existing methods and achieving performance close to traditional MTL.
Authors: Xiang Yu, Yayan Chen, Guannan He, Qing Zeng, Yue Qin, Meiling Liang, Dandan Luo, Yimei Liao, Zeyu Ren, Cheng Kang, Delong Yang, Bocheng Liang, Bin Pu, Ying Yuan, Shengli Li
Abstract: While modern segmentation models often prioritize performance over practicality, we advocate a design philosophy prioritizing simplicity and efficiency, and attempted high performance segmentation model design. This paper presents SimpleUNet, a scalable ultra-lightweight medical image segmentation model with three key innovations: (1) A partial feature selection mechanism in skip connections for redundancy reduction while enhancing segmentation performance; (2) A fixed-width architecture that prevents exponential parameter growth across network stages; (3) An adaptive feature fusion module achieving enhanced representation with minimal computational overhead. With a record-breaking 16 KB parameter configuration, SimpleUNet outperforms LBUNet and other lightweight benchmarks across multiple public datasets. The 0.67 MB variant achieves superior efficiency (8.60 GFLOPs) and accuracy, attaining a mean DSC/IoU of 85.76%/75.60% on multi-center breast lesion datasets, surpassing both U-Net and TransUNet. Evaluations on skin lesion datasets (ISIC 2017/2018: mDice 84.86%/88.77%) and endoscopic polyp segmentation (KVASIR-SEG: 86.46%/76.48% mDice/mIoU) confirm consistent dominance over state-of-the-art models. This work demonstrates that extreme model compression need not compromise performance, providing new insights for efficient and accurate medical image segmentation. Codes can be found at https://github.com/Frankyu5666666/SimpleUNet.
Authors: Daniel Dager, Robin Sobczyk, Emmanuel Chemla, Yair Lakretz
Abstract: It takes several years for the developing brain of a baby to fully master word repetition-the task of hearing a word and repeating it aloud. Repeating a new word, such as from a new language, can be a challenging task also for adults. Additionally, brain damage, such as from a stroke, may lead to systematic speech errors with specific characteristics dependent on the location of the brain damage. Cognitive sciences suggest a model with various components for the different processing stages involved in word repetition. While some studies have begun to localize the corresponding regions in the brain, the neural mechanisms and how exactly the brain performs word repetition remain largely unknown. We propose to bridge the gap between the cognitive model of word repetition and neural mechanisms in the human brain by modeling the task using deep neural networks. Neural models are fully observable, allowing us to study the detailed mechanisms in their various substructures and make comparisons with human behavior and, ultimately, the brain. Here, we make first steps in this direction by: (1) training a large set of models to simulate the word repetition task; (2) creating a battery of tests to probe the models for known effects from behavioral studies in humans, and (3) simulating brain damage through ablation studies, where we systematically remove neurons from the model, and repeat the behavioral study to examine the resulting speech errors in the "patient" model. Our results show that neural models can mimic several effects known from human research, but might diverge in other aspects, highlighting both the potential and the challenges for future research aimed at developing human-like neural models.
Authors: YR Darr, MA Niazi
Abstract: The self-organization of robots for the formation of structures and shapes is a stimulating application of the swarm robotic system. It involves a large number of autonomous robots of heterogeneous behavior, coordination among them, and their interaction with the dynamic environment. This process of complex structure formation is considered a complex system, which needs to be modeled by using any modeling approach. Although the formal specification approach along with other formal methods has been used to model the behavior of robots in a swarm. However, to the best of our knowledge, the formal specification approach has not been used to model the self-organization process in swarm robotic systems for shape formation. In this paper, we use a formal specification approach to model the shape formation task of swarm robots. We use Z (Zed) language of formal specification, which is a state-based language, to model the states of the entities of the systems. We demonstrate the effectiveness of Z for the self-organized shape formation. The presented formal specification model gives the outlines for designing and implementing the swarm robotic system for the formation of complex shapes and structures. It also provides the foundation for modeling the complex shape formation process for swarm robotics using a multi-agent system in a simulation-based environment. Keywords: Swarm robotics, Self-organization, Formal specification, Complex systems
Authors: Zhengyu Hu, Jianxun Lian, Zheyuan Xiao, Seraphina Zhang, Tianfu Wang, Nicholas Jing Yuan, Xing Xie, Hui Xiong
Abstract: Large language models (LLMs) have shown impressive capabilities across tasks such as mathematics, coding, and reasoning, yet their learning ability, which is crucial for adapting to dynamic environments and acquiring new knowledge, remains underexplored. In this work, we address this gap by introducing a framework inspired by cognitive psychology and education. Specifically, we decompose general learning ability into three distinct, complementary dimensions: Learning from Instructor (acquiring knowledge via explicit guidance), Learning from Concept (internalizing abstract structures and generalizing to new contexts), and Learning from Experience (adapting through accumulated exploration and feedback). We conduct a comprehensive empirical study across the three learning dimensions and identify several insightful findings, such as (i) interaction improves learning; (ii) conceptual understanding is scale-emergent and benefits larger models; and (iii) LLMs are effective few-shot learners but not many-shot learners. Based on our framework and empirical findings, we introduce a benchmark that provides a unified and realistic evaluation of LLMs' general learning abilities across three learning cognition dimensions. It enables diagnostic insights and supports evaluation and development of more adaptive and human-like models.
Authors: Marine Carpuat, Omri Asscher, Kalika Bali, Luisa Bentivogli, Fr\'ed\'eric Blain, Lynne Bowker, Monojit Choudhury, Hal Daum\'e III, Kevin Duh, Ge Gao, Alvin Grissom II, Marzena Karpinska, Elaine C. Khoong, William D. Lewis, Andr\'e F. T. Martins, Mary Nurminen, Douglas W. Oard, Maja Popovic, Michel Simard, Fran\c{c}ois Yvon
Abstract: Machine Translation (MT) tools are widely used today, often in contexts where professional translators are not present. Despite progress in MT technology, a gap persists between system development and real-world usage, particularly for non-expert users who may struggle to assess translation reliability. This paper advocates for a human-centered approach to MT, emphasizing the alignment of system design with diverse communicative goals and contexts of use. We survey the literature in Translation Studies and Human-Computer Interaction to recontextualize MT evaluation and design to address the diverse real-world scenarios in which MT is used today.
Authors: Shiqian Guo, Jianqing Liu, Thinh Le, Huaiyu Dai
Abstract: Quantum magnetic sensing based on spin systems has emerged as a new paradigm for detecting ultra-weak magnetic fields with unprecedented sensitivity, revitalizing applications in navigation, geo-localization, biology, and beyond. At the heart of quantum magnetic sensing, from the protocol perspective, lies the design of optimal sensing parameters to manifest and then estimate the underlying signals of interest (SoI). Existing studies on this front mainly rely on adaptive algorithms based on black-box AI models or formula-driven principled searches. However, when the SoI spans a wide range and the quantum sensor has physical constraints, these methods may fail to converge efficiently or optimally, resulting in prolonged interrogation times and reduced sensing accuracy. In this work, we report the design of a new protocol using a two-stage optimization method. In the 1st Stage, a Bayesian neural network with a fixed set of sensing parameters is used to narrow the range of SoI. In the 2nd Stage, a federated reinforcement learning agent is designed to fine-tune the sensing parameters within a reduced search space. The proposed protocol is developed and evaluated in a challenging context of single-shot readout of an NV-center electron spin under a constrained total sensing time budget; and yet it achieves significant improvements in both accuracy and resource efficiency for wide-range D.C. magnetic field estimation compared to the state of the art.
Authors: Junho Yoon, Geom Lee, Donghyeon Jeon, Inho Kang, Seung-Hoon Na
Abstract: Quantization has been widely studied as an effective technique for reducing the memory requirement of large language models (LLMs), potentially improving the latency time as well. Utilizing the characteristic of rotational invariance of transformer, we propose the rotation-based saliency-aware weight quantization (ROSAQ), which identifies salient channels in the projection feature space, not in the original feature space, where the projected "principal" dimensions are naturally considered as "salient" features. The proposed ROSAQ consists of 1) PCA-based projection, which first performs principal component analysis (PCA) on a calibration set and transforms via the PCA projection, 2) Salient channel dentification, which selects dimensions corresponding to the K-largest eigenvalues as salient channels, and 3) Saliency-aware quantization with mixed-precision, which uses FP16 for salient dimensions and INT3/4 for other dimensions. Experiment results show that ROSAQ shows improvements over the baseline saliency-aware quantization on the original feature space and other existing quantization methods. With kernel fusion, ROSAQ presents about 2.3x speed up over FP16 implementation in generating 256 tokens with a batch size of 64.
Authors: David Bani-Harouni, Chantal Pellegrini, Ege \"Ozsoy, Matthias Keicher, Nassir Navab
Abstract: Clinical decision-making is a dynamic, interactive, and cyclic process where doctors have to repeatedly decide on which clinical action to perform and consider newly uncovered information for diagnosis and treatment. Large Language Models (LLMs) have the potential to support clinicians in this process, however, most applications of LLMs in clinical decision support suffer from one of two limitations: Either they assume the unrealistic scenario of immediate availability of all patient information and do not model the interactive and iterative investigation process, or they restrict themselves to the limited "out-of-the-box" capabilities of large pre-trained models without performing task-specific training. In contrast to this, we propose to model clinical decision-making for diagnosis with a hypothesis-driven uncertainty-aware language agent, LA-CDM, that converges towards a diagnosis via repeatedly requesting and interpreting relevant tests. Using a hybrid training paradigm combining supervised and reinforcement learning, we train LA-CDM with three objectives targeting critical aspects of clinical decision-making: accurate hypothesis generation, hypothesis uncertainty estimation, and efficient decision-making. We evaluate our methodology on MIMIC-CDM, a real-world dataset covering four abdominal diseases containing various clinical tests and show the benefit of explicitly training clinical decision-making for increasing diagnostic performance and efficiency.
Authors: Xiem HoangVan, Dang Bui Dinh, Thanh Nguyen Canh, Van-Truong Nguyen
Abstract: Printed Circuit Boards (PCBs) are critical components in modern electronics, which require stringent quality control to ensure proper functionality. However, the detection of defects in small-scale PCBs images poses significant challenges as a result of the low resolution of the captured images, leading to potential confusion between defects and noise. To overcome these challenges, this paper proposes a novel framework, named ESRPCB (edgeguided super-resolution for PCBs defect detection), which combines edgeguided super-resolution with ensemble learning to enhance PCBs defect detection. The framework leverages the edge information to guide the EDSR (Enhanced Deep Super-Resolution) model with a novel ResCat (Residual Concatenation) structure, enabling it to reconstruct high-resolution images from small PCBs inputs. By incorporating edge features, the super-resolution process preserves critical structural details, ensuring that tiny defects remain distinguishable in the enhanced image. Following this, a multi-modal defect detection model employs ensemble learning to analyze the super-resolved
Authors: Mei-Yen Chen, Thi Thu Uyen Hoang, Michael Hahn, M. Saquib Sarfraz
Abstract: Merging or routing low-rank adapters (LoRAs) has emerged as a popular solution for enhancing large language models, particularly when data access is restricted by regulatory or domain-specific constraints. This position paper argues that the research community should shift its focus from developing new merging or routing algorithms to understanding the conditions under which reusing LoRAs is truly effective. Through theoretical analysis and synthetic two-hop reasoning and math word-problem tasks, we examine whether reusing LoRAs enables genuine compositional generalization or merely reflects shallow pattern matching. Evaluating two data-agnostic methods--parameter averaging and dynamic adapter selection--we found that reusing LoRAs often fails to logically integrate knowledge across disjoint fine-tuning datasets, especially when such knowledge is underrepresented during pretraining. Our empirical results, supported by theoretical insights into LoRA's limited expressiveness, highlight the preconditions and constraints of reusing them for unseen tasks and cast doubt on its feasibility as a truly data-free approach. We advocate for pausing the pursuit of novel methods for recycling LoRAs and emphasize the need for rigorous mechanisms to guide future academic research in adapter-based model merging and practical system designs for practitioners.
Authors: Vasiliki Balaska, Ioannis Tsampikos Papapetros, Katerina Maria Oikonomou, Loukas Bampis, Antonios Gasteratos
Abstract: The mining sector increasingly adopts digital tools to improve operational efficiency, safety, and data-driven decision-making. One of the key challenges remains the reliable acquisition of high-resolution, geo-referenced spatial information to support core activities such as extraction planning and on-site monitoring. This work presents an integrated system architecture that combines UAV-based sensing, LiDAR terrain modeling, and deep learning-based object detection to generate spatially accurate information for open-pit mining environments. The proposed pipeline includes geo-referencing, 3D reconstruction, and object localization, enabling structured spatial outputs to be integrated into an industrial digital twin platform. Unlike traditional static surveying methods, the system offers higher coverage and automation potential, with modular components suitable for deployment in real-world industrial contexts. While the current implementation operates in post-flight batch mode, it lays the foundation for real-time extensions. The system contributes to the development of AI-enhanced remote sensing in mining by demonstrating a scalable and field-validated geospatial data workflow that supports situational awareness and infrastructure safety.
Authors: YuQing Xie, Ameya Daigavane, Mit Kotak, Tess Smidt
Abstract: $E(3)$-equivariant neural networks have demonstrated success across a wide range of 3D modelling tasks. A fundamental operation in these networks is the tensor product, which interacts two geometric features in an equivariant manner to create new features. Due to the high computational complexity of the tensor product, significant effort has been invested to optimize the runtime of this operation. For example, Luo et al. (2024) recently proposed the Gaunt tensor product (GTP) which promises a significant speedup. In this work, we provide a careful, systematic analysis of a number of tensor product operations. In particular, we emphasize that different tensor products are not performing the same operation. The reported speedups typically come at the cost of expressivity. We introduce measures of expressivity and interactability to characterize these differences. In addition, we realized the original implementation of GTP can be greatly simplified by directly using a spherical grid at no cost in asymptotic runtime. This spherical grid approach is faster on our benchmarks and in actual training of the MACE interatomic potential by 30\%. Finally, we provide the first systematic microbenchmarks of the various tensor product operations. We find that the theoretical runtime guarantees can differ wildly from empirical performance, demonstrating the need for careful application-specific benchmarking. Code is available at \href{https://github.com/atomicarchitects/PriceofFreedom}{https://github.com/atomicarchitects/PriceofFreedom}
URLs: https://github.com/atomicarchitects/PriceofFreedom, https://github.com/atomicarchitects/PriceofFreedom
Authors: Jie Chen, Hongling Chen, Jinghuai Gao, Chuangji Meng, Tao Yang, XinXin Liang
Abstract: Seismic acoustic impedance plays a crucial role in lithological identification and subsurface structure interpretation. However, due to the inherently ill-posed nature of the inversion problem, directly estimating impedance from post-stack seismic data remains highly challenging. Recently, diffusion models have shown great potential in addressing such inverse problems due to their strong prior learning and generative capabilities. Nevertheless, most existing methods operate in the pixel domain and require multiple iterations, limiting their applicability to field data. To alleviate these limitations, we propose a novel seismic acoustic impedance inversion framework based on a conditional latent generative diffusion model, where the inversion process is made in latent space. To avoid introducing additional training overhead when embedding conditional inputs, we design a lightweight wavelet-based module into the framework to project seismic data and reuse an encoder trained on impedance to embed low-frequency impedance into the latent space. Furthermore, we propose a model-driven sampling strategy during the inversion process of this framework to enhance accuracy and reduce the number of required diffusion steps. Numerical experiments on a synthetic model demonstrate that the proposed method achieves high inversion accuracy and strong generalization capability within only a few diffusion steps. Moreover, application to field data reveals enhanced geological detail and higher consistency with well-log measurements, validating the effectiveness and practicality of the proposed approach.
Authors: Settaluri Lakshmi Sravanthi, Kishan Maharaj, Sravani Gunnu, Abhijit Mishra, Pushpak Bhattacharyya
Abstract: Pragmatics, the ability to infer meaning beyond literal interpretation, is crucial for social cognition and communication. While LLMs have been benchmarked for their pragmatic understanding, improving their performance remains underexplored. Existing methods rely on annotated labels but overlook the reasoning process humans naturally use to interpret implicit meaning. To bridge this gap, we introduce a novel pragmatic dataset, ImpliedMeaningPreference, that includes explicit reasoning (thoughts) for both correct and incorrect interpretations. Through preference-tuning and supervised fine-tuning, we demonstrate that thought-based learning significantly enhances LLMs' pragmatic understanding, improving accuracy by 11.12% across model families. We further discuss a transfer-learning study where we evaluate the performance of thought-based training for the other tasks of pragmatics (presupposition, deixis) that are not seen during the training time and observe an improvement of 16.10% compared to label-trained models.
Authors: Jonathan Hoss, Felix Schelling, Noah Klarmann
Abstract: The classical Job Shop Scheduling Problem (JSSP) focuses on optimizing makespan under deterministic constraints. Real-world production environments introduce additional complexities that cause traditional scheduling approaches to be less effective. Reinforcement learning (RL) holds potential in addressing these challenges, as it allows agents to learn adaptive scheduling strategies. However, there is a lack of a comprehensive, general-purpose frameworks for effectively training and evaluating RL agents under real-world constraints. To address this gap, we propose a modular framework that extends classical JSSP formulations by incorporating key real-world constraints inherent to the shopfloor, including transport logistics, buffer management, machine breakdowns, setup times, and stochastic processing conditions, while also supporting multi-objective optimization. The framework is a customizable solution that offers flexibility in defining problem instances and configuring simulation parameters, enabling adaptation to diverse production scenarios. A standardized interface ensures compatibility with various RL approaches, providing a robust environment for training RL agents and facilitating the standardized comparison of different scheduling methods under dynamic and uncertain conditions. We release JobShopLab as an open-source tool for both research and industrial applications, accessible at: https://github.com/proto-lab-ro/jobshoplab
Authors: Andrew Zhang, Anushka Sivakumar, Chiawei Tang, Chris Thomas
Abstract: Discrete diffusion models are a new class of text generators that offer advantages such as bidirectional context use, parallelizable generation, and flexible prompting compared to autoregressive models. However, a critical limitation of discrete diffusion models is their inability to perform flexible-length or flexible-position text infilling without access to ground-truth positional data. We introduce \textbf{DDOT} (\textbf{D}iscrete \textbf{D}iffusion with \textbf{O}ptimal \textbf{T}ransport Position Coupling), the first discrete diffusion model to overcome this challenge. DDOT jointly denoises token values and token positions, employing a novel sample-level Optimal Transport (OT) coupling. This coupling preserves relative token ordering while dynamically adjusting the positions and length of infilled segments, a capability previously missing in text diffusion. Our method is orthogonal to existing discrete text diffusion methods and is compatible with various pretrained text denoisers. Extensive experiments on text infilling benchmarks such as One-Billion-Word and Yelp demonstrate that DDOT outperforms naive diffusion baselines. Furthermore, DDOT achieves performance on par with state-of-the-art non-autoregressive models and enables significant improvements in training efficiency and flexibility.
Authors: Bernhard Hilpert, Muhan Hou, Kim Baraka, Joost Broekens
Abstract: Reinforcement Learning (RL) agents often exhibit learning behaviors that are not intuitively interpretable by human observers, which can result in suboptimal feedback in collaborative teaching settings. Yet, how humans perceive and interpret RL agent's learning behavior is largely unknown. In a bottom-up approach with two experiments, this work provides a data-driven understanding of the factors of human observers' understanding of the agent's learning process. A novel, observation-based paradigm to directly assess human inferences about agent learning was developed. In an exploratory interview study (\textit{N}=9), we identify four core themes in human interpretations: Agent Goals, Knowledge, Decision Making, and Learning Mechanisms. A second confirmatory study (\textit{N}=34) applied an expanded version of the paradigm across two tasks (navigation/manipulation) and two RL algorithms (tabular/function approximation). Analyses of 816 responses confirmed the reliability of the paradigm and refined the thematic framework, revealing how these themes evolve over time and interrelate. Our findings provide a human-centered understanding of how people make sense of agent learning, offering actionable insights for designing interpretable RL systems and improving transparency in Human-Robot Interaction.
Authors: Yuwei Du, Jie Feng, Jian Yuan, Yong Li
Abstract: Human mobility simulation plays a crucial role in various real-world applications. Recently, to address the limitations of traditional data-driven approaches, researchers have explored leveraging the commonsense knowledge and reasoning capabilities of large language models (LLMs) to accelerate human mobility simulation. However, these methods suffer from several critical shortcomings, including inadequate modeling of urban spaces and poor integration with both individual mobility patterns and collective mobility distributions. To address these challenges, we propose \textbf{C}ityGPT-Powered \textbf{A}gentic framework for \textbf{M}obility \textbf{S}imulation (\textbf{CAMS}), an agentic framework that leverages the language based urban foundation model to simulate human mobility in urban space. \textbf{CAMS} comprises three core modules, including MobExtractor to extract template mobility patterns and synthesize new ones based on user profiles, GeoGenerator to generate anchor points considering collective knowledge and generate candidate urban geospatial knowledge using an enhanced version of CityGPT, TrajEnhancer to retrieve spatial knowledge based on mobility patterns and generate trajectories with real trajectory preference alignment via DPO. Experiments on real-world datasets show that \textbf{CAMS} achieves superior performance without relying on externally provided geospatial information. Moreover, by holistically modeling both individual mobility patterns and collective mobility constraints, \textbf{CAMS} generates more realistic and plausible trajectories. In general, \textbf{CAMS} establishes a new paradigm that integrates the agentic framework with urban-knowledgeable LLMs for human mobility simulation.
Authors: Javad Enayati, Pedram Asef, Alexandre Benoit
Abstract: This paper introduces a novel hybrid AI method combining H filtering and an adaptive linear neuron network for flicker component estimation in power distribution systems.The proposed method leverages the robustness of the H filter to extract the voltage envelope under uncertain and noisy conditions followed by the use of ADALINE to accurately identify flicker frequencies embedded in the envelope.This synergy enables efficient time domain estimation with rapid convergence and noise resilience addressing key limitations of existing frequency domain approaches.Unlike conventional techniques this hybrid AI model handles complex power disturbances without prior knowledge of noise characteristics or extensive training.To validate the method performance we conduct simulation studies based on IEC Standard 61000 4 15 supported by statistical analysis Monte Carlo simulations and real world data.Results demonstrate superior accuracy robustness and reduced computational load compared to Fast Fourier Transform and Discrete Wavelet Transform based estimators.
Authors: Zhiqiang Li, Haiyong Bao, Menghong Guan, Hao Pan, Cheng Huang, Hong-Ning Dai
Abstract: Despite federated learning (FL)'s potential in collaborative learning, its performance has deteriorated due to the data heterogeneity of distributed users. Recently, clustered federated learning (CFL) has emerged to address this challenge by partitioning users into clusters according to their similarity. However, CFL faces difficulties in training when users are unwilling to share their cluster identities due to privacy concerns. To address these issues, we present an innovative Efficient and Robust Secure Aggregation scheme for CFL, dubbed EBS-CFL. The proposed EBS-CFL supports effectively training CFL while maintaining users' cluster identity confidentially. Moreover, it detects potential poisonous attacks without compromising individual client gradients by discarding negatively correlated gradients and aggregating positively correlated ones using a weighted approach. The server also authenticates correct gradient encoding by clients. EBS-CFL has high efficiency with client-side overhead O(ml + m^2) for communication and O(m^2l) for computation, where m is the number of cluster identities, and l is the gradient size. When m = 1, EBS-CFL's computational efficiency of client is at least O(log n) times better than comparison schemes, where n is the number of clients.In addition, we validate the scheme through extensive experiments. Finally, we theoretically prove the scheme's security.
Authors: Francesco Fabbri, Martino Andrea Scarpolini, Angelo Iollo, Francesco Viola, Francesco Tudisco
Abstract: Synthetic data generation plays a crucial role in medical research by mitigating privacy concerns and enabling large-scale patient data analysis. This study presents a beta-Variational Autoencoder Graph Convolutional Neural Network framework for generating synthetic Abdominal Aorta Aneurysms (AAA). Using a small real-world dataset, our approach extracts key anatomical features and captures complex statistical relationships within a compact disentangled latent space. To address data limitations, low-impact data augmentation based on Procrustes analysis was employed, preserving anatomical integrity. The generation strategies, both deterministic and stochastic, manage to enhance data diversity while ensuring realism. Compared to PCA-based approaches, our model performs more robustly on unseen data by capturing complex, nonlinear anatomical variations. This enables more comprehensive clinical and statistical analyses than the original dataset alone. The resulting synthetic AAA dataset preserves patient privacy while providing a scalable foundation for medical research, device testing, and computational modeling.
Authors: Zhiyi Shi, Binjie Wang, Chongjie Si, Yichen Wu, Junsik Kim, Hanspeter Pfister
Abstract: Model editing aims to efficiently update a pre-trained model's knowledge without the need for time-consuming full retraining. While existing pioneering editing methods achieve promising results, they primarily focus on editing single-modal language models (LLMs). However, for vision-language models (VLMs), which involve multiple modalities, the role and impact of each modality on editing performance remain largely unexplored. To address this gap, we explore the impact of textual and visual modalities on model editing and find that: (1) textual and visual representations reach peak sensitivity at different layers, reflecting their varying importance; and (2) editing both modalities can efficiently update knowledge, but this comes at the cost of compromising the model's original capabilities. Based on our findings, we propose DualEdit, an editor that modifies both textual and visual modalities at their respective key layers. Additionally, we introduce a gating module within the more sensitive textual modality, allowing DualEdit to efficiently update new knowledge while preserving the model's original information. We evaluate DualEdit across multiple VLM backbones and benchmark datasets, demonstrating its superiority over state-of-the-art VLM editing baselines as well as adapted LLM editing methods on different evaluation metrics.
Authors: Shulin Tian, Ruiqi Wang, Hongming Guo, Penghao Wu, Yuhao Dong, Xiuying Wang, Jingkang Yang, Hao Zhang, Hongyuan Zhu, Ziwei Liu
Abstract: We introduce Ego-R1, a novel framework for reasoning over ultra-long (i.e., in days and weeks) egocentric videos, which leverages a structured Chain-of-Tool-Thought (CoTT) process, orchestrated by an Ego-R1 Agent trained via reinforcement learning (RL). Inspired by human problem-solving strategies, CoTT decomposes complex reasoning into modular steps, with the RL agent invoking specific tools, one per step, to iteratively and collaboratively answer sub-questions tackling such tasks as temporal retrieval and multi-modal understanding. We design a two-stage training paradigm involving supervised finetuning (SFT) of a pretrained language model using CoTT data and RL to enable our agent to dynamically propose step-by-step tools for long-range reasoning. To facilitate training, we construct a dataset called Ego-R1 Data, which consists of Ego-CoTT-25K for SFT and Ego-QA-4.4K for RL. Furthermore, our Ego-R1 agent is evaluated on a newly curated week-long video QA benchmark, Ego-R1 Bench, which contains human-verified QA pairs from hybrid sources. Extensive results demonstrate that the dynamic, tool-augmented chain-of-thought reasoning by our Ego-R1 Agent can effectively tackle the unique challenges of understanding ultra-long egocentric videos, significantly extending the time coverage from few hours to a week.
Authors: Junfeng Fang, Zijun Yao, Ruipeng Wang, Haokai Ma, Xiang Wang, Tat-Seng Chua
Abstract: The development of large language models (LLMs) has entered in a experience-driven era, flagged by the emergence of environment feedback-driven learning via reinforcement learning and tool-using agents. This encourages the emergenece of model context protocol (MCP), which defines the standard on how should a LLM interact with external services, such as \api and data. However, as MCP becomes the de facto standard for LLM agent systems, it also introduces new safety risks. In particular, MCP introduces third-party services, which are not controlled by the LLM developers, into the agent systems. These third-party MCP services provider are potentially malicious and have the economic incentives to exploit vulnerabilities and sabotage user-agent interactions. In this position paper, we advocate the research community in LLM safety to pay close attention to the new safety risks issues introduced by MCP, and develop new techniques to build safe MCP-powered agent systems. To establish our position, we argue with three key parts. (1) We first construct \framework, a controlled framework to examine safety issues in MCP-powered agent systems. (2) We then conduct a series of pilot experiments to demonstrate the safety risks in MCP-powered agent systems is a real threat and its defense is not trivial. (3) Finally, we give our outlook by showing a roadmap to build safe MCP-powered agent systems. In particular, we would call for researchers to persue the following research directions: red teaming, MCP safe LLM development, MCP safety evaluation, MCP safety data accumulation, MCP service safeguard, and MCP safe ecosystem construction. We hope this position paper can raise the awareness of the research community in MCP safety and encourage more researchers to join this important research direction. Our code is available at https://github.com/littlelittlenine/SafeMCP.git.
Authors: Haonan Wang, Brian Chen, Siquan Li, Xinhe Liang, Hwee Kuan Lee, Kenji Kawaguchi, Tianyang Hu
Abstract: Parameter-Efficient Fine-Tuning (PEFT) methods have become crucial for rapidly adapting large language models (LLMs) to downstream tasks. Prefix-Tuning, an early and effective PEFT technique, demonstrated the ability to achieve performance comparable to full fine-tuning with significantly reduced computational and memory overhead. However, despite its earlier success, its effectiveness in training modern state-of-the-art LLMs has been very limited. In this work, we demonstrate empirically that Prefix-Tuning underperforms on LLMs because of an inherent tradeoff between input and prefix significance within the attention head. This motivates us to introduce Prefix-Tuning+, a novel architecture that generalizes the principles of Prefix-Tuning while addressing its shortcomings by shifting the prefix module out of the attention head itself. We further provide an overview of our construction process to guide future users when constructing their own context-based methods. Our experiments show that, across a diverse set of benchmarks, Prefix-Tuning+ consistently outperforms existing Prefix-Tuning methods. Notably, it achieves performance on par with the widely adopted LoRA method on several general benchmarks, highlighting the potential modern extension of Prefix-Tuning approaches. Our findings suggest that by overcoming its inherent limitations, Prefix-Tuning can remain a competitive and relevant research direction in the landscape of parameter-efficient LLM adaptation.
Authors: Yuqing Wen, Kefan Gu, Haoxuan Liu, Yucheng Zhao, Tiancai Wang, Haoqiang Fan, Xiaoyan Sun
Abstract: Vision-Language-Action (VLA) models have recently made significant advance in multi-task, end-to-end robotic control, due to the strong generalization capabilities of Vision-Language Models (VLMs). A fundamental challenge in developing such models is effectively aligning the vision-language space with the robotic action space. Existing approaches typically rely on directly fine-tuning VLMs using expert demonstrations. However, this strategy suffers from a spatio-temporal gap, resulting in considerable data inefficiency and heavy reliance on human labor. Spatially, VLMs operate within a high-level semantic space, whereas robotic actions are grounded in low-level 3D physical space; temporally, VLMs primarily interpret the present, while VLA models anticipate future actions. To overcome these challenges, we propose a novel training paradigm, ROSA, which leverages robot state estimation to improve alignment between vision-language and action spaces. By integrating robot state estimation data obtained via an automated process, ROSA enables the VLA model to gain enhanced spatial understanding and self-awareness, thereby boosting performance and generalization. Extensive experiments in both simulated and real-world environments demonstrate the effectiveness of ROSA, particularly in low-data regimes.
Authors: Ionel-Alexandru Hosu, Traian Rebedea, Razvan Pascanu
Abstract: One proposed mechanism to improve exploration in reinforcement learning is through the use of macro-actions. Paradoxically though, in many scenarios the naive addition of macro-actions does not lead to better exploration, but rather the opposite. It has been argued that this was caused by adding non-useful macros and multiple works have focused on mechanisms to discover effectively environment-specific useful macros. In this work, we take a slightly different perspective. We argue that the difficulty stems from the trade-offs between reducing the average number of decisions per episode versus increasing the size of the action space. Namely, one typically treats each potential macro-action as independent and atomic, hence strictly increasing the search space and making typical exploration strategies inefficient. To address this problem we propose a novel regularization term that exploits the relationship between actions and macro-actions to improve the credit assignment mechanism by reducing the effective dimension of the action space and, therefore, improving exploration. The term relies on a similarity matrix that is meta-learned jointly with learning the desired policy. We empirically validate our strategy looking at macro-actions in Atari games, and the StreetFighter II environment. Our results show significant improvements over the Rainbow-DQN baseline in all environments. Additionally, we show that the macro-action similarity is transferable to related environments. We believe this work is a small but important step towards understanding how the similarity-imposed geometry on the action space can be exploited to improve credit assignment and exploration, therefore making learning more effective.
Authors: Shang-Chi Tsai, Yun-Nung Chen
Abstract: With the advancement of large language models, many dialogue systems are now capable of providing reasonable and informative responses to patients' medical conditions. However, when patients consult their doctor, they may experience negative emotions due to the severity and urgency of their situation. If the model can provide appropriate comfort and empathy based on the patient's negative emotions while answering medical questions, it will likely offer a more reassuring experience during the medical consultation process. To address this issue, our paper explores the balance between knowledge sharing and emotional support in the healthcare dialogue process. We utilize a large language model to rewrite a real-world interactive medical dialogue dataset, generating patient queries with negative emotions and corresponding medical responses aimed at soothing the patient's emotions while addressing their concerns. The modified data serves to refine the latest large language models with various fine-tuning methods, enabling them to accurately provide sentences with both emotional reassurance and constructive suggestions in response to patients' questions. Compared to the original LLM model, our experimental results demonstrate that our methodology significantly enhances the model's ability to generate emotional responses while maintaining its original capability to provide accurate knowledge-based answers.
Authors: Bilal Faye, Hanane Azzag, Mustapha Lebbah
Abstract: Single-trajectory reinforcement learning (RL) methods aim to optimize policies from datasets consisting of (prompt, response, reward) triplets, where scalar rewards are directly available. This supervision format is highly practical, as it mirrors real-world human feedback, such as thumbs-up/down signals, and avoids the need for structured preference annotations. In contrast, pairwise preference-based methods like Direct Preference Optimization (DPO) rely on datasets with both preferred and dispreferred responses, which are harder to construct and less natural to collect. Among single-trajectory approaches, Direct Reward Optimization (DRO) has shown strong empirical performance due to its simplicity and stability. However, DRO requires approximating a value function, which introduces several limitations: high off-policy variance, coupling between policy and value learning, and a lack of absolute supervision on the policy itself. We introduce Reward Partitioning Optimization (RPO), a new method that resolves these limitations by removing the need to model the value function. Instead, RPO normalizes observed rewards using a partitioning approach estimated directly from data. This leads to a straightforward supervised learning objective on the policy, with no auxiliary models and no joint optimization. RPO provides direct and stable supervision on the policy, making it robust and easy to implement in practice. We validate RPO on scalar-feedback language modeling tasks using Flan-T5 encoder-decoder models. Our results demonstrate that RPO outperforms existing single-trajectory baselines such as DRO and Kahneman-Tversky Optimization (KTO). These findings confirm that RPO is a simple, effective, and theoretically grounded method for single-trajectory policy optimization.
Authors: Junru Zhang, Lang Feng, Xu Guo, Yuhan Wu, Yabo Dong, Duanqing Xu
Abstract: Time-series reasoning remains a significant challenge in multimodal large language models (MLLMs) due to the dynamic temporal patterns, ambiguous semantics, and lack of temporal priors. In this work, we introduce TimeMaster, a reinforcement learning (RL)-based method that enables time-series MLLMs to perform structured, interpretable reasoning directly over visualized time-series inputs and task prompts. TimeMaster adopts a three-part structured output format, reasoning, classification, and domain-specific extension, and is optimized via a composite reward function that aligns format adherence, prediction accuracy, and open-ended insight quality. The model is trained using a two-stage pipeline: we first apply supervised fine-tuning (SFT) to establish a good initialization, followed by Group Relative Policy Optimization (GRPO) at the token level to enable stable and targeted reward-driven improvement in time-series reasoning. We evaluate TimeMaster on the TimerBed benchmark across six real-world classification tasks based on Qwen2.5-VL-3B-Instruct. TimeMaster achieves state-of-the-art performance, outperforming both classical time-series models and few-shot GPT-4o by over 14.6% and 7.3% performance gain, respectively. Notably, TimeMaster goes beyond time-series classification: it also exhibits expert-like reasoning behavior, generates context-aware explanations, and delivers domain-aligned insights. Our results highlight that reward-driven RL can be a scalable and promising path toward integrating temporal understanding into time-series MLLMs.
Authors: Guanming Zhang, David J. Heeger, Stefano Martiniani
Abstract: Contrastive self-supervised learning based on point-wise comparisons has been widely studied for vision tasks. In the visual cortex of the brain, neuronal responses to distinct stimulus classes are organized into geometric structures known as neural manifolds. Accurate classification of stimuli can be achieved by effectively separating these manifolds, akin to solving a packing problem. We introduce Contrastive Learning As Manifold Packing (CLAMP), a self-supervised framework that recasts representation learning as a manifold packing problem. CLAMP introduces a loss function inspired by the potential energy of short-range repulsive particle systems, such as those encountered in the physics of simple liquids and jammed packings. In this framework, each class consists of sub-manifolds embedding multiple augmented views of a single image. The sizes and positions of the sub-manifolds are dynamically optimized by following the gradient of a packing loss. This approach yields interpretable dynamics in the embedding space that parallel jamming physics, and introduces geometrically meaningful hyperparameters within the loss function. Under the standard linear evaluation protocol, which freezes the backbone and trains only a linear classifier, CLAMP achieves competitive performance with state-of-the-art self-supervised models. Furthermore, our analysis reveals that neural manifolds corresponding to different categories emerge naturally and are effectively separated in the learned representation space, highlighting the potential of CLAMP to bridge insights from physics, neural science, and machine learning.
Authors: Sayed Mohammad Vakilzadeh Hatefi, Maximilian Dreyer, Reduan Achtibat, Patrick Kahardipraja, Thomas Wiegand, Wojciech Samek, Sebastian Lapuschkin
Abstract: Large Language Models (LLMs) are central to many contemporary AI applications, yet their extensive parameter counts pose significant challenges for deployment in memory- and compute-constrained environments. Recent works in eXplainable AI (XAI), particularly on attribution methods, suggest that interpretability can also enable model compression by identifying and removing components irrelevant to inference. In this paper, we leverage Layer-wise Relevance Propagation (LRP) to perform attribution-guided pruning of LLMs. While LRP has shown promise in structured pruning for vision models, we extend it to unstructured pruning in LLMs and demonstrate that it can substantially reduce model size with minimal performance loss. Our method is especially effective in extracting task-relevant subgraphs -- so-called ``circuits'' -- which can represent core functions (e.g., indirect object identification). Building on this, we introduce a technique for model correction, by selectively removing circuits responsible for spurious behaviors (e.g., toxic outputs). All in all, we gather these techniques as a uniform holistic framework and showcase its effectiveness and limitations through extensive experiments for compression, circuit discovery and model correction on Llama and OPT models, highlighting its potential for improving both model efficiency and safety. Our code is publicly available at https://github.com/erfanhatefi/SparC3.
Authors: Tain\~a Coleman, Hena Ahmed, Ravi Shende, Ismael Perez, \"Ilkay Altinta\c{s}
Abstract: Distributed computing systems are essential for meeting the demands of modern applications, yet transitioning from single-system to distributed environments presents significant challenges. Misallocating resources in shared systems can lead to resource contention, system instability, degraded performance, priority inversion, inefficient utilization, increased latency, and environmental impact. We present BanditWare, an online recommendation system that dynamically selects the most suitable hardware for applications using a contextual multi-armed bandit algorithm. BanditWare balances exploration and exploitation, gradually refining its hardware recommendations based on observed application performance while continuing to explore potentially better options. Unlike traditional statistical and machine learning approaches that rely heavily on large historical datasets, BanditWare operates online, learning and adapting in real-time as new workloads arrive. We evaluated BanditWare on three workflow applications: Cycles (an agricultural science scientific workflow) BurnPro3D (a web-based platform for fire science) and a matrix multiplication application. Designed for seamless integration with the National Data Platform (NDP), BanditWare enables users of all experience levels to optimize resource allocation efficiently.
Authors: Vitoria Guardieiro, Adam Stein, Avishree Khare, Eric Wong
Abstract: Controlling the generation of large language models (LLMs) remains a central challenge to ensure their safe and reliable deployment. While prompt engineering and finetuning are common approaches, recent work has explored latent steering, a lightweight technique that alters LLM internal activations to guide generation. However, subsequent studies revealed latent steering's effectiveness to be limited, often underperforming simple instruction prompting. To address this limitation, we first establish a benchmark across diverse behaviors for standardized evaluation of steering techniques. Building on insights from this benchmark, we introduce Instruction Attention Boosting (InstABoost), a latent steering method that boosts the strength of instruction prompting by altering the model's attention during generation. InstABoost combines the strengths of existing approaches and is theoretically supported by prior work that suggests that in-context rule following in transformer-based models can be controlled by manipulating attention on instructions. Empirically, InstABoost demonstrates superior control success compared to both traditional prompting and latent steering.
Authors: Shova Kuikel, Aritran Piplai, Palvi Aggarwal
Abstract: Phishing attacks remain one of the most prevalent and persistent cybersecurity threat with attackers continuously evolving and intensifying tactics to evade the general detection system. Despite significant advances in artificial intelligence and machine learning, faithfully reproducing the interpretable reasoning with classification and explainability that underpin phishing judgments remains challenging. Due to recent advancement in Natural Language Processing, Large Language Models (LLMs) show a promising direction and potential for improving domain specific phishing classification tasks. However, enhancing the reliability and robustness of classification models requires not only accurate predictions from LLMs but also consistent and trustworthy explanations aligning with those predictions. Therefore, a key question remains: can LLMs not only classify phishing emails accurately but also generate explanations that are reliably aligned with their predictions and internally self-consistent? To answer these questions, we have fine-tuned transformer based models, including BERT, Llama models, and Wizard, to improve domain relevance and make them more tailored to phishing specific distinctions, using Binary Sequence Classification, Contrastive Learning (CL) and Direct Preference Optimization (DPO). To that end, we examined their performance in phishing classification and explainability by applying the ConsistenCy measure based on SHAPley values (CC SHAP), which measures prediction explanation token alignment to test the model's internal faithfulness and consistency and uncover the rationale behind its predictions and reasoning. Overall, our findings show that Llama models exhibit stronger prediction explanation token alignment with higher CC SHAP scores despite lacking reliable decision making accuracy, whereas Wizard achieves better prediction accuracy but lower CC SHAP scores.
Authors: Haoru Xue, Xiaoyu Huang, Dantong Niu, Qiayuan Liao, Thomas Kragerud, Jan Tommy Gravdahl, Xue Bin Peng, Guanya Shi, Trevor Darrell, Koushil Screenath, Shankar Sastry
Abstract: Vision-language-action (VLA) models have demonstrated strong semantic understanding and zero-shot generalization, yet most existing systems assume an accurate low-level controller with hand-crafted action "vocabulary" such as end-effector pose or root velocity. This assumption confines prior work to quasi-static tasks and precludes the agile, whole-body behaviors required by humanoid whole-body control (WBC) tasks. To capture this gap in the literature, we start by introducing the first sim-to-real-ready, vision-language, closed-loop benchmark for humanoid WBC, comprising over 150 tasks from 10 categories. We then propose LeVERB: Latent Vision-Language-Encoded Robot Behavior, a hierarchical latent instruction-following framework for humanoid vision-language WBC, the first of its kind. At the top level, a vision-language policy learns a latent action vocabulary from synthetically rendered kinematic demonstrations; at the low level, a reinforcement-learned WBC policy consumes these latent verbs to generate dynamics-level commands. In our benchmark, LeVERB can zero-shot attain a 80% success rate on simple visual navigation tasks, and 58.5% success rate overall, outperforming naive hierarchical whole-body VLA implementation by 7.8 times.
Authors: Junyan Li, Wenshuo Zhao, Yang Zhang, Chuang Gan
Abstract: Recent deep-thinking large language models often reason extensively to improve performance, but such lengthy reasoning is not always desirable, as it incurs excessive inference costs with disproportionate performance gains. Controlling reasoning length without sacrificing performance is therefore important, but remains challenging, especially under tight thinking budgets. We propose budget guidance, a simple yet effective method for steering the reasoning process of LLMs toward a target budget without requiring any LLM fine-tuning. Our approach introduces a lightweight predictor that models a Gamma distribution over the remaining thinking length during next-token generation. This signal is then used to guide generation in a soft, token-level manner, ensuring that the overall reasoning trace adheres to the specified thinking budget. Budget guidance enables natural control of the thinking length, along with significant token efficiency improvements over baseline methods on challenging math benchmarks. For instance, it achieves up to a 26% accuracy gain on the MATH-500 benchmark under tight budgets compared to baseline methods, while maintaining competitive accuracy with only 63% of the thinking tokens used by the full-thinking model. Budget guidance also generalizes to broader task domains and exhibits emergent capabilities, such as estimating question difficulty. The source code is available at: https://github.com/UMass-Embodied-AGI/BudgetGuidance.
Authors: Edward Li, Zichen Wang, Jiahe Huang, Jeong Joon Park
Abstract: We present a unified framework for solving partial differential equations (PDEs) using video-inpainting diffusion transformer models. Unlike existing methods that devise specialized strategies for either forward or inverse problems under full or partial observation, our approach unifies these tasks under a single, flexible generative framework. Specifically, we recast PDE-solving as a generalized inpainting problem, e.g., treating forward prediction as inferring missing spatiotemporal information of future states from initial conditions. To this end, we design a transformer-based architecture that conditions on arbitrary patterns of known data to infer missing values across time and space. Our method proposes pixel-space video diffusion models for fine-grained, high-fidelity inpainting and conditioning, while enhancing computational efficiency through hierarchical modeling. Extensive experiments show that our video inpainting-based diffusion model offers an accurate and versatile solution across a wide range of PDEs and problem setups, outperforming state-of-the-art baselines.
Authors: Runpeng Yu, Qi Li, Xinchao Wang
Abstract: In this work, we provide a systematic survey of Discrete Diffusion Language Models (dLLMs) and Discrete Diffusion Multimodal Language Models (dMLLMs). Unlike autoregressive (AR) models, dLLMs and dMLLMs adopt a multi-token, parallel decoding paradigm using full attention and a denoising-based generation strategy. This paradigm naturally enables parallel generation, fine-grained output controllability, and dynamic, response-aware perception. These capabilities are previously difficult to achieve with AR models. Recently, a growing number of industrial-scale proprietary d(M)LLMs, as well as a large number of open-source academic d(M)LLMs, have demonstrated performance comparable to their autoregressive counterparts, while achieving up to 10x acceleration in inference speed. The advancement of discrete diffusion LLMs and MLLMs has been largely driven by progress in two domains. The first is the development of autoregressive LLMs and MLLMs, which has accumulated vast amounts of data, benchmarks, and foundational infrastructure for training and inference. The second contributing domain is the evolution of the mathematical models underlying discrete diffusion. Together, these advancements have catalyzed a surge in dLLMs and dMLLMs research in early 2025. In this work, we present a comprehensive overview of the research in the dLLM and dMLLM domains. We trace the historical development of dLLMs and dMLLMs, formalize the underlying mathematical frameworks, and categorize representative models. We further analyze key techniques for training and inference, and summarize emerging applications across language, vision-language, and biological domains. We conclude by discussing future directions for research and deployment. Paper collection: https://github.com/LiQiiiii/DLLM-Survey
Authors: Yixian Xu, Shengjie Luo, Liwei Wang, Di He, Chang Liu
Abstract: Diffusion models have achieved remarkable success in generative modeling. Despite more stable training, the loss of diffusion models is not indicative of absolute data-fitting quality, since its optimal value is typically not zero but unknown, leading to confusion between large optimal loss and insufficient model capacity. In this work, we advocate the need to estimate the optimal loss value for diagnosing and improving diffusion models. We first derive the optimal loss in closed form under a unified formulation of diffusion models, and develop effective estimators for it, including a stochastic variant scalable to large datasets with proper control of variance and bias. With this tool, we unlock the inherent metric for diagnosing the training quality of mainstream diffusion model variants, and develop a more performant training schedule based on the optimal loss. Moreover, using models with 120M to 1.5B parameters, we find that the power law is better demonstrated after subtracting the optimal loss from the actual training loss, suggesting a more principled setting for investigating the scaling law for diffusion models.
Authors: Martin Schmalzried
Abstract: This paper leverages various philosophical and ontological frameworks to explore the concept of embodied artificial general intelligence (AGI), its relationship to human consciousness, and the key role of the metaverse in facilitating this relationship. Several theoretical frameworks underpin this exploration, such as embodied cognition, Michael Levin's computational boundary of a "Self," and Donald D. Hoffman's Interface Theory of Perception, which lead to considering human perceived outer reality as a symbolic representation of alternate inner states of being, and where AGI could embody a different form of consciousness with a larger computational boundary. The paper further discusses the necessary architecture for the emergence of an embodied AGI, how to calibrate an AGI's symbolic interface, and the key role played by the Metaverse, decentralized systems and open-source blockchain technology. The paper concludes by emphasizing the importance of achieving a certain degree of harmony in human relations and recognizing the interconnectedness of humanity at a global level, as key prerequisites for the emergence of a stable embodied AGI.
Authors: Zhaowei Zhang, Fengshuo Bai, Mingzhi Wang, Haoyang Ye, Chengdong Ma, Yaodong Yang
Abstract: The burgeoning integration of artificial intelligence (AI) into human society brings forth significant implications for societal governance and safety. While considerable strides have been made in addressing AI alignment challenges, existing methodologies primarily focus on technical facets, often neglecting the intricate sociotechnical nature of AI systems, which can lead to a misalignment between the development and deployment contexts. To this end, we posit a new problem worth exploring: Incentive Compatibility Sociotechnical Alignment Problem (ICSAP). We hope this can call for more researchers to explore how to leverage the principles of Incentive Compatibility (IC) from game theory to bridge the gap between technical and societal components to maintain AI consensus with human societies in different contexts. We further discuss three classical game problems for achieving IC: mechanism design, contract theory, and Bayesian persuasion, in addressing the perspectives, potentials, and challenges of solving ICSAP, and provide preliminary implementation conceptions.
Authors: Jiaming Ji, Donghai Hong, Borong Zhang, Boyuan Chen, Juntao Dai, Boren Zheng, Tianyi Qiu, Jiayi Zhou, Kaile Wang, Boxuan Li, Sirui Han, Yike Guo, Yaodong Yang
Abstract: In this study, we introduce the safety human preference dataset, PKU-SafeRLHF, designed to promote research on safety alignment in large language models (LLMs). As a sibling project to SafeRLHF and BeaverTails, we separate annotations of helpfulness and harmlessness for question-answering pairs, providing distinct perspectives on these coupled attributes. Overall, we provide 44.6k refined prompts and 265k question-answer pairs with safety meta-labels for 19 harm categories and three severity levels ranging from minor to severe, with answers generated by Llama-family models. Based on this, we collected 166.8k preference data, including dual-preference (helpfulness and harmlessness decoupled) and single-preference data (trade-off the helpfulness and harmlessness from scratch), respectively. Using the large-scale annotation data, we further train severity-sensitive moderation for the risk control of LLMs and safety-centric RLHF algorithms for the safety alignment of LLMs. We believe this dataset will be a valuable resource for the community, aiding in the safe deployment of LLMs. Data is available at https://huggingface.co/datasets/PKU-Alignment/PKU-SafeRLHF.
URLs: https://huggingface.co/datasets/PKU-Alignment/PKU-SafeRLHF.
Authors: Yoshua Bengio, Michael K. Cohen, Nikolay Malkin, Matt MacDermott, Damiano Fornasiere, Pietro Greiner, Younesse Kaddar
Abstract: Is there a way to design powerful AI systems based on machine learning methods that would satisfy probabilistic safety guarantees? With the long-term goal of obtaining a probabilistic guarantee that would apply in every context, we consider estimating a context-dependent bound on the probability of violating a given safety specification. Such a risk evaluation would need to be performed at run-time to provide a guardrail against dangerous actions of an AI. Noting that different plausible hypotheses about the world could produce very different outcomes, and because we do not know which one is right, we derive bounds on the safety violation probability predicted under the true but unknown hypothesis. Such bounds could be used to reject potentially dangerous actions. Our main results involve searching for cautious but plausible hypotheses, obtained by a maximization that involves Bayesian posteriors over hypotheses. We consider two forms of this result, in the i.i.d. case and in the non-i.i.d. case, and conclude with open problems towards turning such theoretical results into practical AI guardrails.
Authors: Yuhao Jia, Zile Wu, Shengao Yi, Yifei Sun, Xiao Huang
Abstract: Urban forecasting has increasingly benefited from high-dimensional spatial data through two primary approaches: graph-based methods that rely on predefined spatial structures, and region-based methods that focus on learning expressive urban representations. Although these methods have laid a strong foundation, they either rely heavily on structured spatial data, struggle to adapt to task-specific dependencies, or fail to integrate holistic urban context. Moreover, no existing framework systematically integrates these two paradigms and overcomes their respective limitations. To address this gap, we propose a novel, unified framework for high-dimensional urban forecasting, composed of three key components: (1) the Urban Region Representation Module that organizes latent embeddings and semantic descriptions for each region, (2) the Task-aware Dependency Retrieval module that selects relevant context regions based on natural language prompts, and (3) the Prediction Module, exemplified by our proposed GeoTransformer architecture, which adopts a novel geospatial attention mechanism to incorporate spatial proximity and information entropy as priors. Our framework is modular, supports diverse representation methods and forecasting models, and can operate even with minimal input. Quantitative experiments and qualitative analysis across six urban forecasting tasks demonstrate strong task generalization and validate the framework's effectiveness.
Authors: Haolong Chen, Hanzhi Chen, Zijian Zhao, Kaifeng Han, Guangxu Zhu, Yichen Zhao, Ying Du, Wei Xu, Qingjiang Shi
Abstract: The impressive performance of ChatGPT and other foundation-model-based products in human language understanding has prompted both academia and industry to explore how these models can be tailored for specific industries and application scenarios. This process, known as the customization of domain-specific foundation models (FMs), addresses the limitations of general-purpose models, which may not fully capture the unique patterns and requirements of domain-specific data. Despite its importance, there is a notable lack of comprehensive overview papers on building domain-specific FMs, while numerous resources exist for general-purpose models. To bridge this gap, this article provides a timely and thorough overview of the methodology for customizing domain-specific FMs. It introduces basic concepts, outlines the general architecture, and surveys key methods for constructing domain-specific models. Furthermore, the article discusses various domains that can benefit from these specialized models and highlights the challenges ahead. Through this overview, we aim to offer valuable guidance and reference for researchers and practitioners from diverse fields to develop their own customized FMs.
Authors: Dongwei Jiang, Guoxuan Wang, Yining Lu, Andrew Wang, Jingyu Zhang, Chuyu Liu, Benjamin Van Durme, Daniel Khashabi
Abstract: The reasoning steps generated by LLMs might be incomplete, as they mimic logical leaps common in everyday communication found in their pre-training data: underlying rationales are frequently left implicit (unstated). To address this challenge, we introduce RATIONALYST, a model for process-supervision of reasoning based on pre-training on a vast collection of rationale annotations extracted from unlabeled data. We extract 79k rationales from web-scale unlabelled dataset (the Pile) and a combination of reasoning datasets with minimal human intervention. This web-scale pre-training for reasoning allows RATIONALYST to consistently generalize across diverse reasoning tasks, including mathematical, commonsense, scientific, and logical reasoning. Fine-tuned from LLaMa-3-8B, RATIONALYST improves the accuracy of reasoning by an average of 3.9% on 7 representative reasoning benchmarks. It also demonstrates superior performance compared to significantly larger verifiers like GPT-4 and similarly sized models fine-tuned on matching training sets.
Authors: Jingyu Peng, Maolin Wang, Xiangyu Zhao, Kai Zhang, Wanyu Wang, Pengyue Jia, Qidong Liu, Ruocheng Guo, Qi Liu
Abstract: Large language models (LLMs) have made remarkable strides in complex reasoning tasks, but their safety and robustness in reasoning processes remain underexplored. Existing attacks on LLM reasoning are constrained by specific settings or lack of imperceptibility, limiting their feasibility and generalizability. To address these challenges, we propose the Stepwise rEasoning Error Disruption (SEED) attack, which subtly injects errors into prior reasoning steps to mislead the model into producing incorrect subsequent reasoning and final answers. Unlike previous methods, SEED is compatible with zero-shot and few-shot settings, maintains the natural reasoning flow, and ensures covert execution without modifying the instruction. Extensive experiments on four datasets across four different models demonstrate SEED's effectiveness, revealing the vulnerabilities of LLMs to disruptions in reasoning processes. These findings underscore the need for greater attention to the robustness of LLM reasoning to ensure safety in practical applications. Our code is available at: https://github.com/Applied-Machine-Learning-Lab/SEED-Attack.
URLs: https://github.com/Applied-Machine-Learning-Lab/SEED-Attack.
Authors: Jio Oh, Geon Heo, Seungjun Oh, Hyunjin Kim, JinYeong Bak, Jindong Wang, Xing Xie, Steven Euijong Whang
Abstract: Large Language Models (LLMs) often struggle with data-analytics requests related to information retrieval and data manipulation that frequently arise in real-world scenarios under multiple conditions. In this paper, we introduce Thinking with Tables, where we inject tabular structures into LLMs for data-analytics requests. Through comprehensive evaluations across various request types, we show that providing tabular structures yields a 40.29 percent average performance gain along with better robustness and token efficiency. Through attention-value analysis, we uncover that tables help LLMs better attend to relevant information, explaining these improvements. Beyond tables and text, we evaluate whether (1) blending structuredness within text, such as providing templates or fixing the order of attributes, and (2) other representative structures, such as knowledge graphs and JSON, are helpful. We observe that utilizing tables offers the best balance between efficiency and effectiveness. These advantages remain consistent under increased task complexity and even when all input data cannot be structured. Finally, as data analytics typically relies on structured factual inputs, our text-to-table conversion demonstrates the method's applicability to text-compatible data sources.
Authors: Yong Lai, Haolong Tong, Zhenghang Xu, Minghao Yin
Abstract: Quantitative information flow analyses (QIF) are a class of techniques for measuring the amount of confidential information leaked by a program to its public outputs. Shannon entropy is an important method to quantify the amount of leakage in QIF. This paper focuses on the programs modeled in Boolean constraints and optimizes the two stages of the Shannon entropy computation to implement a scalable precise tool PSE. In the first stage, we design a knowledge compilation language called \ADDAND that combines Algebraic Decision Diagrams and conjunctive decomposition. \ADDAND avoids enumerating possible outputs of a program and supports tractable entropy computation. In the second stage, we optimize the model counting queries that are used to compute the probabilities of outputs. We compare PSE with the state-of-the-art probabilistic approximately correct tool EntropyEstimation, which was shown to significantly outperform the previous precise tools. The experimental results demonstrate that PSE solved 56 more benchmarks compared to EntropyEstimation in a total of 459. For 98\% of the benchmarks that both PSE and EntropyEstimation solved, PSE is at least $10\times$ as efficient as EntropyEstimation.
Authors: Luis A. Pineda
Abstract: Artificial computing machinery transforms representations through an objective process, to be interpreted subjectively by humans, so the machine and the interpreter are different entities, but in the putative natural computing both processes are performed by the same agent. The method or process that transforms a representation is called here the mode of computing. The mode used by digital computers is the algorithmic one, but there are others, such as quantum computers and diverse forms of non-conventional computing, and there is an open-ended set of representational formats and modes that could be used in artificial and natural computing. A mode based on a notion of computing different from Turing's may perform feats beyond what the Turing Machine does but the modes would not be of the same kind and could not be compared. For a mode of computing to be more powerful than the algorithmic one, it ought to compute functions lacking an effective algorithm, and Church Thesis would not hold. Here, a thought experiment including a computational demon using a hypothetical mode for such an effect is presented. If there is natural computing, there is a mode of natural computing whose properties may be causal to the phenomenological experience. Discovering it would come with solving the hard problem of consciousness; but if it turns out that such a mode does not exist, there is no such thing as natural computing, and the mind is not a computational process.
Authors: Zhong-Zhi Li, Duzhen Zhang, Ming-Liang Zhang, Jiaxin Zhang, Zengyan Liu, Yuxuan Yao, Haotian Xu, Junhao Zheng, Pei-Jie Wang, Xiuyi Chen, Yingying Zhang, Fei Yin, Jiahua Dong, Zhiwei Li, Bao-Long Bi, Ling-Rui Mei, Junfeng Fang, Xiao Liang, Zhijiang Guo, Le Song, Cheng-Lin Liu
Abstract: Achieving human-level intelligence requires refining the transition from the fast, intuitive System 1 to the slower, more deliberate System 2 reasoning. While System 1 excels in quick, heuristic decisions, System 2 relies on logical reasoning for more accurate judgments and reduced biases. Foundational Large Language Models (LLMs) excel at fast decision-making but lack the depth for complex reasoning, as they have not yet fully embraced the step-by-step analysis characteristic of true System 2 thinking. Recently, reasoning LLMs like OpenAI's o1/o3 and DeepSeek's R1 have demonstrated expert-level performance in fields such as mathematics and coding, closely mimicking the deliberate reasoning of System 2 and showcasing human-like cognitive abilities. This survey begins with a brief overview of the progress in foundational LLMs and the early development of System 2 technologies, exploring how their combination has paved the way for reasoning LLMs. Next, we discuss how to construct reasoning LLMs, analyzing their features, the core methods enabling advanced reasoning, and the evolution of various reasoning LLMs. Additionally, we provide an overview of reasoning benchmarks, offering an in-depth comparison of the performance of representative reasoning LLMs. Finally, we explore promising directions for advancing reasoning LLMs and maintain a real-time \href{https://github.com/zzli2022/Awesome-Slow-Reason-System}{GitHub Repository} to track the latest developments. We hope this survey will serve as a valuable resource to inspire innovation and drive progress in this rapidly evolving field.
URLs: https://github.com/zzli2022/Awesome-Slow-Reason-System
Authors: Nayoung Choi, Grace Byun, Andrew Chung, Ellie S. Paek, Shinsun Lee, Jinho D. Choi
Abstract: Proprietary corporate documents contain rich domain-specific knowledge, but their overwhelming volume and disorganized structure make it difficult even for employees to access the right information when needed. For example, in the automotive industry, vehicle crash-collision tests, each costing hundreds of thousands of dollars, produce highly detailed documentation. However, retrieving relevant content during decision-making remains time-consuming due to the scale and complexity of the material. While Retrieval-Augmented Generation (RAG)-based Question Answering (QA) systems offer a promising solution, building an internal RAG-QA system poses several challenges: (1) handling heterogeneous multi-modal data sources, (2) preserving data confidentiality, and (3) enabling traceability between each piece of information in the generated answer and its original source document. To address these, we propose a RAG-QA framework for internal enterprise use, consisting of: (1) a data pipeline that converts raw multi-modal documents into a structured corpus and QA pairs, (2) a fully on-premise, privacy-preserving architecture, and (3) a lightweight reference matcher that links answer segments to supporting content. Applied to the automotive domain, our system improves factual correctness (+1.79, +1.94), informativeness (+1.33, +1.16), and helpfulness (+1.08, +1.67) over a non-RAG baseline, based on 1-5 scale ratings from both human and LLM judge.
Authors: Shraddha Surana, Ashwin Srinivasan
Abstract: We are interested in the construction of software that can act as scientific assistants to domain specialists. It is expected that such assistants will be needed to accelerate the identification of ways to address complex problems requiring urgent solutions. In this paper, our focus is not on a specific scientific problem, but on the software-engineering of such 'science accelerators'. Recent developments in 'No Code' techniques would seem to suggest that scientist can simply hypothesise solutions simply by conversing with a large language model (LLM). However, for complex scientific problems, this seems unlikely given the current state of LLM technology. What does appear feasible is that a software engineer can use LLMs to rapidly construct programs for use by a domain-specialist, including the specialist's requirements expressed in natural language. We propose the design of an interactive form of 'structured' inductive programming in which a software-engineer and an LLM collaboratively construct an 'assistant' for a scientific data analysis. The paper describes a simple implementation called iStrucInd that adapts a '2-way Intelligibility' protocol to implement the interaction between the software engineer and the LLM. We test the tool on two different non-trivial scientific data analysis tasks. Specifically, we compare the system constructed by iStrucInd against systems constructed manually and by Low Code/No Code methods along dimensions of: (a) program performance; (b) program quality; and (c) programming effort. The results show iStrucInd allows a software engineer to develop better programs faster suggesting interactive structured induction can play a useful role in the rapid construction of scientific assistants.
Authors: Maciej Besta, Lorenzo Paleari, Jia Hao Andrea Jiang, Robert Gerstenberger, You Wu, J\'on Gunnar Hannesson, Patrick Iff, Ales Kubicek, Piotr Nyczyk, Diana Khimey, Nils Blach, Haiqiang Zhang, Tao Zhang, Peiran Ma, Grzegorz Kwa\'sniewski, Marcin Copik, Hubert Niewiadomski, Torsten Hoefler
Abstract: Large Language Models (LLMs) are revolutionizing the development of AI assistants capable of performing diverse tasks across domains. However, current state-of-the-art LLM-driven agents face significant challenges, including high operational costs and limited success rates on complex benchmarks like GAIA. To address these issues, we propose Knowledge Graph of Thoughts (KGoT), an innovative AI assistant architecture that integrates LLM reasoning with dynamically constructed knowledge graphs (KGs). KGoT extracts and structures task-relevant knowledge into a dynamic KG representation, iteratively enhanced through external tools such as math solvers, web crawlers, and Python scripts. Such structured representation of task-relevant knowledge enables low-cost models to solve complex tasks effectively while also minimizing bias and noise. For example, KGoT achieves a 29% improvement in task success rates on the GAIA benchmark compared to Hugging Face Agents with GPT-4o mini. Moreover, harnessing a smaller model dramatically reduces operational costs by over 36x compared to GPT-4o. Improvements for other models (e.g., Qwen2.5-32B and Deepseek-R1-70B) and benchmarks (e.g., SimpleQA) are similar. KGoT offers a scalable, affordable, versatile, and high-performing solution for AI assistants.
Authors: Robert E. Wray, Steven J. Jones, John E. Laird
Abstract: Regardless of past learning, an agent in an open world will face unfamiliar events outside of prior experience, existing models, or policies. Further, the agent will sometimes lack relevant knowledge and/or sufficient time to assess the situation and evaluate response options. How can an agent respond reasonably to situations that are outside of its original design scope? How can it recognize such situations sufficiently quickly and reliably to determine reasonable, adaptive courses of action? We identify key characteristics needed for solutions, review the state-of-the-art, and outline a proposed, novel approach that combines domain-general meta-knowledge (inspired by human cognition) and metareasoning. This approach offers potential for fast, adaptive responses to unfamiliar situations, more fully meeting the performance characteristics required for open-world, general agents.
Authors: Robert E. Wray, James R. Kirk, John E. Laird
Abstract: One goal of AI (and AGI) is to identify and understand specific mechanisms and representations sufficient for general intelligence. Often, this work manifests in research focused on architectures and many cognitive architectures have been explored in AI/AGI. However, different research groups and even different research traditions have somewhat independently identified similar/common patterns of processes and representations or "cognitive design patterns" that are manifest in existing architectures. Today, AI systems exploiting large language models (LLMs) offer a relatively new combination of mechanisms and representations available for exploring the possibilities of general intelligence. This paper outlines a few recurring cognitive design patterns that have appeared in various pre-transformer AI architectures. We then explore how these patterns are evident in systems using LLMs, especially for reasoning and interactive ("agentic") use cases. Examining and applying these recurring patterns enables predictions of gaps or deficiencies in today's Agentic LLM Systems and identification of subjects of future research towards general intelligence using generative foundation models.
Authors: Hao Li, He Cao, Bin Feng, Yanjun Shao, Xiangru Tang, Zhiyuan Yan, Li Yuan, Yonghong Tian, Yu Li
Abstract: While large language models (LLMs) with Chain-of-Thought (CoT) reasoning excel in mathematics and coding, their potential for systematic reasoning in chemistry, a domain demanding rigorous structural analysis for real-world tasks like drug design and reaction engineering, remains untapped. Current benchmarks focus on simple knowledge retrieval, neglecting step-by-step reasoning required for complex tasks such as molecular optimization and reaction prediction. To address this, we introduce ChemCoTBench, a reasoning framework that bridges molecular structure understanding with arithmetic-inspired operations, including addition, deletion, and substitution, to formalize chemical problem-solving into transparent, step-by-step workflows. By treating molecular transformations as modular "chemical operations", the framework enables slow-thinking reasoning, mirroring the logic of mathematical proofs while grounding solutions in real-world chemical constraints. We evaluate models on two high-impact tasks: Molecular Property Optimization and Chemical Reaction Prediction. These tasks mirror real-world challenges while providing structured evaluability. By providing annotated datasets, a reasoning taxonomy, and baseline evaluations, ChemCoTBench bridges the gap between abstract reasoning methods and practical chemical discovery, establishing a foundation for advancing LLMs as tools for AI-driven scientific innovation.
Authors: Mingyang Mao, Mariela M. Perez-Cabarcas, Utteja Kallakuri, Nicholas R. Waytowich, Xiaomin Lin, Tinoosh Mohsenin
Abstract: To effectively engage in human society, the ability to adapt, filter information, and make informed decisions in ever-changing situations is critical. As robots and intelligent agents become more integrated into human life, there is a growing opportunity-and need-to offload the cognitive burden on humans to these systems, particularly in dynamic, information-rich scenarios. To fill this critical need, we present Multi-RAG, a multimodal retrieval-augmented generation system designed to provide adaptive assistance to humans in information-intensive circumstances. Our system aims to improve situational understanding and reduce cognitive load by integrating and reasoning over multi-source information streams, including video, audio, and text. As an enabling step toward long-term human-robot partnerships, Multi-RAG explores how multimodal information understanding can serve as a foundation for adaptive robotic assistance in dynamic, human-centered situations. To evaluate its capability in a realistic human-assistance proxy task, we benchmarked Multi-RAG on the MMBench-Video dataset, a challenging multimodal video understanding benchmark. Our system achieves superior performance compared to existing open-source video large language models (Video-LLMs) and large vision-language models (LVLMs), while utilizing fewer resources and less input data. The results demonstrate Multi- RAG's potential as a practical and efficient foundation for future human-robot adaptive assistance systems in dynamic, real-world contexts.
Authors: Prashik Buddhaghosh Bansod
Abstract: The emergence of large language models has catalyzed two distinct yet interconnected paradigms in artificial intelligence: standalone AI Agents and collaborative Agentic AI ecosystems. This comprehensive study establishes a definitive framework for distinguishing these architectures through systematic analysis of their operational principles, structural compositions, and deployment methodologies. We characterize AI Agents as specialized, tool-enhanced systems leveraging foundation models for targeted automation within constrained environments. Conversely, Agentic AI represents sophisticated multi-entity frameworks where distributed agents exhibit emergent collective intelligence through coordinated interaction protocols. Our investigation traces the evolutionary trajectory from traditional rule-based systems through generative AI foundations to contemporary agent architectures. We present detailed architectural comparisons examining planning mechanisms, memory systems, coordination protocols, and decision-making processes. The study categorizes application landscapes, contrasting single-agent implementations in customer service and content management with multi-agent deployments in research automation and complex decision support. We identify critical challenges including reliability issues, coordination complexities, and scalability constraints, while proposing innovative solutions through enhanced reasoning frameworks, robust memory architectures, and improved coordination mechanisms. This framework provides essential guidance for practitioners selecting appropriate agentic approaches and establishes foundational principles for next-generation intelligent system development.
Authors: Jonathan Richens, David Abel, Alexis Bellot, Tom Everitt
Abstract: Are world models a necessary ingredient for flexible, goal-directed behaviour, or is model-free learning sufficient? We provide a formal answer to this question, showing that any agent capable of generalizing to multi-step goal-directed tasks must have learned a predictive model of its environment. We show that this model can be extracted from the agent's policy, and that increasing the agents performance or the complexity of the goals it can achieve requires learning increasingly accurate world models. This has a number of consequences: from developing safe and general agents, to bounding agent capabilities in complex environments, and providing new algorithms for eliciting world models from agents.
Authors: Mikhail Terekhov, Zhen Ning David Liu, Caglar Gulcehre, Samuel Albanie
Abstract: The rapid integration of agentic AI into high-stakes real-world applications requires robust oversight mechanisms. The emerging field of AI Control (AIC) aims to provide such an oversight mechanism, but practical adoption depends heavily on implementation overhead. To study this problem better, we introduce the notion of Control tax -- the operational and financial cost of integrating control measures into AI pipelines. Our work makes three key contributions to the field of AIC: (1) we introduce a theoretical framework that quantifies the Control Tax and maps classifier performance to safety assurances; (2) we conduct comprehensive evaluations of state-of-the-art language models in adversarial settings, where attacker models insert subtle backdoors into code while monitoring models attempt to detect these vulnerabilities; and (3) we provide empirical financial cost estimates for control protocols and develop optimized monitoring strategies that balance safety and cost-effectiveness while accounting for practical constraints like auditing budgets. Our framework enables practitioners to make informed decisions by systematically connecting safety guarantees with their costs, advancing AIC through principled economic feasibility assessment across different deployment contexts.
Authors: John Beverley, Jim Logan, Barry Smith
Abstract: This paper introduces a framework for representing information about entities that do not exist or may never exist, such as those involving fictional entities, blueprints, simulations, and future scenarios. Traditional approaches that introduce "dummy instances" or rely on modal logic are criticized, and a proposal is defended in which such cases are modeled using the intersections of actual types rather than specific non existent tokens. The paper positions itself within the Basic Formal Ontology and its realist commitments, emphasizing the importance of practical, implementable solutions over purely metaphysical or philosophical proposals, arguing that existing approaches to non existent entities either overcommit to metaphysical assumptions or introduce computational inefficiencies that hinder applications. By developing a structured ontology driven approach to unreal patterns, the paper aims to provide a useful and computationally viable means of handling references to hypothetical or non existent entities.
Authors: Cameron Angliss, Jiaxun Cui, Jiaheng Hu, Arrasy Rahman, Peter Stone
Abstract: Developing AI agents that can robustly adapt to dramatically different strategic landscapes without retraining is a central challenge for multi-agent learning. Pok\'emon Video Game Championships (VGC) is a domain with an extraordinarily large space of possible team configurations of approximately $10^{139}$ - far larger than those of Dota or Starcraft. The highly discrete, combinatorial nature of team building in Pok\'emon VGC causes optimal strategies to shift dramatically depending on both the team being piloted and the opponent's team, making generalization uniquely challenging. To advance research on this problem, we introduce VGC-Bench: a benchmark that provides critical infrastructure, standardizes evaluation protocols, and supplies human-play datasets and a range of baselines - from large-language-model agents and behavior cloning to reinforcement learning and empirical game-theoretic methods such as self-play, fictitious play, and double oracle. In the restricted setting where an agent is trained and evaluated on a single-team configuration, our methods are able to win against a professional VGC competitor. We extensively evaluated all baseline methods over progressively larger team sets and find that even the best-performing algorithm in the single-team setting struggles at scaling up as team size grows. Thus, policy generalization across diverse team strategies remains an open challenge for the community. Our code is open sourced at https://github.com/cameronangliss/VGC-Bench.
Authors: Claudio Fanconi, Mihaela van der Schaar
Abstract: Effective human-AI decision-making balances three key factors: the \textit{correctness} of predictions, the \textit{cost} of knowledge and reasoning complexity, and the confidence about whether to \textit{abstain} automated answers or involve human experts. In this work, we present a cascaded LLM decision framework that adaptively delegates tasks across multiple tiers of expertise -- a base model for initial candidate answers, a more capable and knowledgeable (but costlier) large model, and a human expert for when the model cascade abstains. Our method proceeds in two stages. First, a deferral policy determines whether to accept the base model's answer or regenerate it with the large model based on the confidence score. Second, an abstention policy decides whether the cascade model response is sufficiently certain or requires human intervention. Moreover, we incorporate an online learning mechanism in the framework that can leverage human feedback to improve decision quality over time. We demonstrate this approach to general question-answering (ARC-Easy and ARC-Challenge) and medical question-answering (MedQA and MedMCQA). Our results show that our cascaded strategy outperforms in most cases single-model baselines in accuracy while reducing cost and providing a principled way to handle abstentions.
Authors: Liad Erez, Tal Lancewicki, Uri Sherman, Tomer Koren, Yishay Mansour
Abstract: An abundance of recent impossibility results establish that regret minimization in Markov games with adversarial opponents is both statistically and computationally intractable. Nevertheless, none of these results preclude the possibility of regret minimization under the assumption that all parties adopt the same learning procedure. In this work, we present the first (to our knowledge) algorithm for learning in general-sum Markov games that provides sublinear regret guarantees when executed by all agents. The bounds we obtain are for swap regret, and thus, along the way, imply convergence to a correlated equilibrium. Our algorithm is decentralized, computationally efficient, and does not require any communication between agents. Our key observation is that online learning via policy optimization in Markov games essentially reduces to a form of weighted regret minimization, with unknown weights determined by the path length of the agents' policy sequence. Consequently, controlling the path length leads to weighted regret objectives for which sufficiently adaptive algorithms provide sublinear regret guarantees.
Authors: Anastasios N. Angelopoulos, Stephen Bates, Adam Fisch, Lihua Lei, Tal Schuster
Abstract: We extend conformal prediction to control the expected value of any monotone loss function. The algorithm generalizes split conformal prediction together with its coverage guarantee. Like conformal prediction, the conformal risk control procedure is tight up to an $\mathcal{O}(1/n)$ factor. We also introduce extensions of the idea to distribution shift, quantile risk control, multiple and adversarial risk control, and expectations of U-statistics. Worked examples from computer vision and natural language processing demonstrate the usage of our algorithm to bound the false negative rate, graph distance, and token-level F1-score.
Authors: Siyuan Li, Zedong Wang, Zicheng Liu, Cheng Tan, Haitao Lin, Di Wu, Zhiyuan Chen, Jiangbin Zheng, Stan Z. Li
Abstract: By contextualizing the kernel as global as possible, Modern ConvNets have shown great potential in computer vision tasks. However, recent progress on multi-order game-theoretic interaction within deep neural networks (DNNs) reveals the representation bottleneck of modern ConvNets, where the expressive interactions have not been effectively encoded with the increased kernel size. To tackle this challenge, we propose a new family of modern ConvNets, dubbed MogaNet, for discriminative visual representation learning in pure ConvNet-based models with favorable complexity-performance trade-offs. MogaNet encapsulates conceptually simple yet effective convolutions and gated aggregation into a compact module, where discriminative features are efficiently gathered and contextualized adaptively. MogaNet exhibits great scalability, impressive efficiency of parameters, and competitive performance compared to state-of-the-art ViTs and ConvNets on ImageNet and various downstream vision benchmarks, including COCO object detection, ADE20K semantic segmentation, 2D&3D human pose estimation, and video prediction. Notably, MogaNet hits 80.0% and 87.8% accuracy with 5.2M and 181M parameters on ImageNet-1K, outperforming ParC-Net and ConvNeXt-L, while saving 59% FLOPs and 17M parameters, respectively. The source code is available at https://github.com/Westlake-AI/MogaNet.
Authors: M. Alex O. Vasilescu
Abstract: We derive a set of causal deep neural networks whose architectures are a consequence of tensor (multilinear) factor analysis, a framework that facilitates causal inference. Forward causal questions are addressed with a neural network architecture composed of causal capsules and a tensor transformer. Causal capsules compute a set of invariant causal factor representations, whose interactions are governed by a tensor transformation. Inverse causal questions are addressed with a neural network that implements the multilinear projection algorithm. The architecture reverses the order of operations of a forward neural network and estimates the causes of effects. As an alternative to aggressive bottleneck dimension reduction or regularized regression that may camouflage an inherently underdetermined inverse problem, we prescribe modeling different aspects of the mechanism of data formation with piecewise tensor models whose multilinear projections produce multiple candidate solutions. Our forward and inverse questions may be addressed with shallow architectures, but for computationally scalable solutions, we derive a set of deep neural networks by taking advantage of block algebra. An interleaved kernel hierarchy results in doubly non-linear tensor factor models. The causal neural networks that are a consequence of tensor factor analysis are data agnostic, but are illustrated with facial images. Sequential, parallel and asynchronous parallel computation strategies are described.
Authors: Le Ma, Ran Zhang, Yikun Han, Shirui Yu, Zaitian Wang, Zhiyuan Ning, Jinghan Zhang, Ping Xu, Pengjiang Li, Wei Ju, Chong Chen, Dongjie Wang, Kunpeng Liu, Pengyang Wang, Pengfei Wang, Yanjie Fu, Chunjiang Liu, Yuanchun Zhou, Chang-Tien Lu
Abstract: Vector databases (VDBs) have emerged to manage high-dimensional data that exceed the capabilities of traditional database management systems, and are now tightly integrated with large language models as well as widely applied in modern artificial intelligence systems. Although relatively few studies describe existing or introduce new vector database architectures, the core technologies underlying VDBs, such as approximate nearest neighbor search, have been extensively studied and are well documented in the literature. In this work, we present a comprehensive review of the relevant algorithms to provide a general understanding of this booming research area. Specifically, we first provide a review of storage and retrieval techniques in VDBs, with detailed design principles and technological evolution. Then, we conduct an in-depth comparison of several advanced VDB solutions with their strengths, limitations, and typical application scenarios. Finally, we also outline emerging opportunities for coupling VDBs with large language models, including open research problems and trends, such as novel indexing strategies. This survey aims to serve as a practical resource, enabling readers to quickly gain an overall understanding of the current knowledge landscape in this rapidly developing area.
Authors: Zian Li, Xiyuan Wang, Shijia Kang, Muhan Zhang
Abstract: Invariant models, one important class of geometric deep learning models, are capable of generating meaningful geometric representations by leveraging informative geometric features in point clouds. These models are characterized by their simplicity, good experimental results and computational efficiency. However, their theoretical expressive power still remains unclear, restricting a deeper understanding of the potential of such models. In this work, we concentrate on characterizing the theoretical expressiveness of a wide range of invariant models under fully-connected conditions. We first rigorously characterize the expressiveness of the most classic invariant model, message-passing neural networks incorporating distance (DisGNN), restricting its unidentifiable cases to be only highly symmetric point clouds. We then prove that GeoNGNN, the geometric counterpart of one of the simplest subgraph graph neural networks, can effectively break these corner cases' symmetry and thus achieve E(3)-completeness. By leveraging GeoNGNN as a theoretical tool, we further prove that: 1) most subgraph GNNs developed in traditional graph learning can be seamlessly extended to geometric scenarios with E(3)-completeness; 2) DimeNet, GemNet and SphereNet, three well-established invariant models, are also all capable of achieving E(3)-completeness. Our theoretical results fill the gap in the expressive power of invariant models, contributing to a rigorous and comprehensive understanding of their capabilities.
Authors: Melanie Roschewitz, Fabio De Sousa Ribeiro, Tian Xia, Galvin Khara, Ben Glocker
Abstract: Contrastive pretraining is well-known to improve downstream task performance and model generalisation, especially in limited label settings. However, it is sensitive to the choice of augmentation pipeline. Positive pairs should preserve semantic information while destroying domain-specific information. Standard augmentation pipelines emulate domain-specific changes with pre-defined photometric transformations, but what if we could simulate realistic domain changes instead? In this work, we show how to utilise recent progress in counterfactual image generation to this effect. We propose CF-SimCLR, a counterfactual contrastive learning approach which leverages approximate counterfactual inference for positive pair creation. Comprehensive evaluation across five datasets, on chest radiography and mammography, demonstrates that CF-SimCLR substantially improves robustness to acquisition shift with higher downstream performance on both in- and out-of-distribution data, particularly for domains which are under-represented during training.
Authors: Shuhao Li, Yue Cui, Jingyi Xu, Libin Li, Lingkai Meng, Weidong Yang, Fan Zhang, Xiaofang Zhou
Abstract: Traffic prediction has long been a focal and pivotal area in research, witnessing both significant strides from city-level to road-level predictions in recent years. With the advancement of Vehicle-to-Everything (V2X) technologies, autonomous driving, and large-scale models in the traffic domain, lane-level traffic prediction has emerged as an indispensable direction. However, further progress in this field is hindered by the absence of comprehensive and unified evaluation standards, coupled with limited public availability of data and code. In this paper, we present the first systematic classification framework for lane-level traffic prediction, offering a structured taxonomy and analysis of existing methods. We construct three representative datasets from two real-world road networks, covering both regular and irregular lane configurations, and make them publicly available to support future research. We further establishes a unified spatial topology structure and prediction task formulation, and proposes a simple yet effective baseline model, GraphMLP, based on graph structure and MLP networks. This unified framework enables consistent evaluation across datasets and modeling paradigms. We also reproduce previously unavailable code from existing studies and conduct extensive experiments to assess a range of models in terms of accuracy, efficiency, and applicability, providing the first benchmark that jointly considers predictive performance and training cost for lane-level traffic scenarios. All datasets and code are released at https://github.com/ShuhaoLii/LaneLevel-Traffic-Benchmark.
URLs: https://github.com/ShuhaoLii/LaneLevel-Traffic-Benchmark.
Authors: Feibo Jiang, Li Dong, Siwei Tu, Yubo Peng, Kezhi Wang, Kun Yang, Cunhua Pan, Dusit Niyato
Abstract: Large language models (LLMs) have driven profound transformations in wireless networks. However, within wireless environments, the training of LLMs faces significant challenges related to security and privacy. Federated Learning (FL), with its decentralized architecture, offers enhanced data privacy protection. Nevertheless, when integrated with LLMs, FL still struggles with several critical limitations, including large-scale and heterogeneous data, resource-intensive training, and substantial communication overhead. To address these challenges, this paper first presents a systematic analysis of the distinct training stages of LLMs in wireless networks, including pre-training, instruction tuning, and alignment tuning. Building upon this foundation, we propose a Personalized Wireless Federated Fine-tuning (PWFF) framework. Initially, we utilize the adapter and Low-Rank Adaptation (LoRA) techniques to decrease energy consumption, while employing global partial aggregation to reduce communication delay. Subsequently, we develop two reward models and design a personalized loss function to fulfill the goal of personalized learning. Furthermore, we implement a local multi-objective alignment to ensure the stability and effectiveness of the FL process. Finally, we conduct a series of simulations to validate the performance of the proposed PWFF method and provide an in-depth discussion of the open issues.
Authors: Evan W. Patton, David Y. J. Kim, Ashley Granquist, Robin Liu, Arianna Scott, Jennet Zamanova, Harold Abelson
Abstract: This paper introduces Aptly, a platform designed to democratize mobile app development, particularly for young learners. Aptly integrates a Large Language Model (LLM) with App Inventor, enabling users to create apps using their natural language. User's description is translated into a programming language that corresponds with App Inventor's visual blocks. A preliminary study with high school students demonstrated the usability and potential of the platform. Prior programming experience influenced how users interact with Aptly. Participants identified areas for improvement and expressed a shift in perspective regarding programming accessibility and AI's role in creative endeavors.
Authors: Zijiang Yan, Hina Tabassum
Abstract: We develop a novel multi-objective reinforcement learning (MORL) framework to jointly optimize wireless network selection and autonomous driving policies in a multi-band vehicular network operating on conventional sub-6GHz spectrum and Terahertz frequencies. The proposed framework is designed to 1. maximize the traffic flow and minimize collisions by controlling the vehicle's motion dynamics (i.e., speed and acceleration), and 2. enhance the ultra-reliable low-latency communication (URLLC) while minimizing handoffs (HOs). We cast this problem as a multi-objective Markov Decision Process (MOMDP) and develop solutions for both predefined and unknown preferences of the conflicting objectives. Specifically, we develop a novel envelope MORL solution which develops policies that address multiple objectives with unknown preferences to the agent. While this approach reduces reliance on scalar rewards, policy effectiveness varying with different preferences is a challenge. To address this, we apply a generalized version of the Bellman equation and optimize the convex envelope of multi-objective Q values to learn a unified parametric representation capable of generating optimal policies across all possible preference configurations. Following an initial learning phase, our agent can execute optimal policies under any specified preference or infer preferences from minimal data samples. Numerical results validate the efficacy of the envelope-based MORL solution and demonstrate interesting insights related to the inter-dependency of vehicle motion dynamics, HOs, and the communication data rate. The proposed policies enable autonomous vehicles (AVs) to adopt safe driving behaviors with improved connectivity.
Authors: Jiyuan Tan, Jose Blanchet, Vasilis Syrgkanis
Abstract: Recent progress in Neural Causal Models (NCMs) showcased how identification and partial identification of causal effects can be automatically carried out via training of neural generative models that respect the constraints encoded in a given causal graph [Xia et al. 2022, Balazadeh et al. 2022]. However, formal consistency of these methods has only been proven for the case of discrete variables or only for linear causal models. In this work, we prove the consistency of partial identification via NCMs in a general setting with both continuous and categorical variables. Further, our results highlight the impact of the design of the underlying neural network architecture in terms of depth and connectivity as well as the importance of applying Lipschitz regularization in the training phase. In particular, we provide a counterexample showing that without Lipschitz regularization this method may not be asymptotically consistent. Our results are enabled by new results on the approximability of Structural Causal Models (SCMs) via neural generative models, together with an analysis of the sample complexity of the resulting architectures and how that translates into an error in the constrained optimization problem that defines the partial identification bounds.
Authors: Sarthak Mittal, Eric Elmoznino, Leo Gagnon, Sangnie Bhardwaj, Tom Marty, Dhanya Sridhar, Guillaume Lajoie
Abstract: Large autoregressive models like Transformers can solve tasks through in-context learning (ICL) without learning new weights, suggesting avenues for efficiently solving new tasks. For many tasks, e.g., linear regression, the data factorizes: examples are independent given a task latent that generates the data, e.g., linear coefficients. While an optimal predictor leverages this factorization by inferring task latents, it is unclear if Transformers implicitly do so or if they instead exploit heuristics and statistical shortcuts enabled by attention layers. Both scenarios have inspired active ongoing work. In this paper, we systematically investigate the effect of explicitly inferring task latents. We minimally modify the Transformer architecture with a bottleneck designed to prevent shortcuts in favor of more structured solutions, and then compare performance against standard Transformers across various ICL tasks. Contrary to intuition and some recent works, we find little discernible difference between the two; biasing towards task-relevant latent variables does not lead to better out-of-distribution performance, in general. Curiously, we find that while the bottleneck effectively learns to extract latent task variables from context, downstream processing struggles to utilize them for robust prediction. Our study highlights the intrinsic limitations of Transformers in achieving structured ICL solutions that generalize, and shows that while inferring the right latents aids interpretability, it is not sufficient to alleviate this problem.
Authors: Honghao Fu, Yufei Wang, Wenhan Yang, Alex C. Kot, Bihan Wen
Abstract: Blind image quality assessment (IQA) in the wild, which assesses the quality of images with complex authentic distortions and no reference images, presents significant challenges. Given the difficulty in collecting large-scale training data, leveraging limited data to develop a model with strong generalization remains an open problem. Motivated by the robust image perception capabilities of pre-trained text-to-image (T2I) diffusion models, we propose a novel IQA method, diffusion priors-based IQA (DP-IQA), to utilize the T2I model's prior for improved performance and generalization ability. Specifically, we utilize pre-trained Stable Diffusion as the backbone, extracting multi-level features from the denoising U-Net guided by prompt embeddings through a tunable text adapter. Simultaneously, an image adapter compensates for information loss introduced by the lossy pre-trained encoder. Unlike T2I models that require full image distribution modeling, our approach targets image quality assessment, which inherently requires fewer parameters. To improve applicability, we distill the knowledge into a lightweight CNN-based student model, significantly reducing parameters while maintaining or even enhancing generalization performance. Experimental results demonstrate that DP-IQA achieves state-of-the-art performance on various in-the-wild datasets, highlighting the superior generalization capability of T2I priors in blind IQA tasks. To our knowledge, DP-IQA is the first method to apply pre-trained diffusion priors in blind IQA. Codes and checkpoints are available at https://github.com/RomGai/DP-IQA.
Authors: Justin Cui, Wei-Lin Chiang, Ion Stoica, Cho-Jui Hsieh
Abstract: Large Language Models (LLMs) require careful safety alignment to prevent malicious outputs. While significant research focuses on mitigating harmful content generation, the enhanced safety often come with the side effect of over-refusal, where LLMs may reject innocuous prompts and become less helpful. Although the issue of over-refusal has been empirically observed, a systematic measurement is challenging due to the difficulty of crafting prompts that can elicit the over-refusal behaviors of LLMs. This study proposes a novel method for automatically generating large-scale over-refusal datasets. Leveraging this technique, we introduce OR-Bench, the first large-scale over-refusal benchmark. OR-Bench comprises 80,000 over-refusal prompts across 10 common rejection categories, a subset of around 1,000 hard prompts that are challenging even for state-of-the-art LLMs, and an additional 600 toxic prompts to prevent indiscriminate responses. We then conduct a comprehensive study to measure the over-refusal of 32 popular LLMs across 8 model families. Our datasets are publicly available at https://huggingface.co/bench-llms and our codebase is open-sourced at https://github.com/justincui03/or-bench. We hope this benchmark can help the community develop better safety aligned models.
URLs: https://huggingface.co/bench-llms, https://github.com/justincui03/or-bench.
Authors: Junfeng Jiao, Saleh Afroogh, Yiming Xu, Connor Phillips
Abstract: This study addresses ethical issues surrounding Large Language Models (LLMs) within the field of artificial intelligence. It explores the common ethical challenges posed by both LLMs and other AI systems, such as privacy and fairness, as well as ethical challenges uniquely arising from LLMs. It highlights challenges such as hallucination, verifiable accountability, and decoding censorship complexity, which are unique to LLMs and distinct from those encountered in traditional AI systems. The study underscores the need to tackle these complexities to ensure accountability, reduce biases, and enhance transparency in the influential role that LLMs play in shaping information dissemination. It proposes mitigation strategies and future directions for LLM ethics, advocating for interdisciplinary collaboration. It recommends ethical frameworks tailored to specific domains and dynamic auditing systems adapted to diverse contexts. This roadmap aims to guide responsible development and integration of LLMs, envisioning a future where ethical considerations govern AI advancements in society.
Authors: Vedang Lad, Jin Hwa Lee, Wes Gurnee, Max Tegmark
Abstract: We investigate the robustness of Large Language Models (LLMs) to structural interventions by deleting and swapping adjacent layers during inference. Surprisingly, models retain 72-95% of their original top-1 prediction accuracy without any fine-tuning. We find that performance degradation is not uniform across layers: interventions to the early and final layers cause the most degradation, while the model is remarkably robust to dropping middle layers. This pattern of localized sensitivity motivates our hypothesis of four stages of inference, observed across diverse model families and sizes: (1) detokenization, where local context is integrated to lift raw token embeddings into higher-level representations; (2) feature engineering, where task- and entity-specific features are iteratively refined; (3) prediction ensembling, where hidden states are aggregated into plausible next-token predictions; and (4) residual sharpening, where irrelevant features are suppressed to finalize the output distribution. Synthesizing behavioral and mechanistic evidence, we provide a framework for interpreting depth-dependent computations in LLMs.
Authors: Jintai Chen, Yaojun Hu, Mingchen Cai, Yingzhou Lu, Yue Wang, Xu Cao, Miao Lin, Hongxia Xu, Jian Wu, Cao Xiao, Jimeng Sun, Yuqiang Li, Lucas Glass, Kexin Huang, Marinka Zitnik, Tianfan Fu
Abstract: Clinical trials are pivotal for developing new medical treatments but typically carry risks such as patient mortality and enrollment failure that waste immense efforts spanning over a decade. Applying artificial intelligence (AI) to predict key events in clinical trials holds great potential for providing insights to guide trial designs. However, complex data collection and question definition requiring medical expertise have hindered the involvement of AI thus far. This paper tackles these challenges by presenting a comprehensive suite of 23 meticulously curated AI-ready datasets covering multi-modal input features and 8 crucial prediction challenges in clinical trial design, encompassing prediction of trial duration, patient dropout rate, serious adverse event, mortality rate, trial approval outcome, trial failure reason, drug dose finding, design of eligibility criteria. Furthermore, we provide basic validation methods for each task to ensure the datasets' usability and reliability. We anticipate that the availability of such open-access datasets will catalyze the development of advanced AI approaches for clinical trial design, ultimately advancing clinical trial research and accelerating medical solution development.
Authors: Junbin Xiao, Nanxin Huang, Hangyu Qin, Dongyang Li, Yicong Li, Fengbin Zhu, Zhulin Tao, Jianxing Yu, Liang Lin, Tat-Seng Chua, Angela Yao
Abstract: Video Large Language Models (Video-LLMs) are flourishing and has advanced many video-language tasks. As a golden testbed, Video Question Answering (VideoQA) plays pivotal role in Video-LLM developing. This work conducts a timely and comprehensive study of Video-LLMs' behavior in VideoQA, aiming to elucidate their success and failure modes, and provide insights towards more human-like video understanding and question answering. Our analyses demonstrate that Video-LLMs excel in VideoQA; they can correlate contextual cues and generate plausible responses to questions about varied video contents. However, models falter in handling video temporality, both in reasoning about temporal content ordering and grounding QA-relevant temporal moments. Moreover, the models behave unintuitively - they are unresponsive to adversarial video perturbations while being sensitive to simple variations of candidate answers and questions. Also, they do not necessarily generalize better. The findings demonstrate Video-LLMs' QA capability in standard condition yet highlight their severe deficiency in robustness and interpretability, suggesting the urgent need on rationales in Video-LLM developing.
Authors: Xinyu Liu, Shuyu Shen, Boyan Li, Peixian Ma, Runzhi Jiang, Yuxin Zhang, Ju Fan, Guoliang Li, Nan Tang, Yuyu Luo
Abstract: Translating users' natural language queries (NL) into SQL queries (i.e., Text-to-SQL, a.k.a. NL2SQL) can significantly reduce barriers to accessing relational databases and support various commercial applications. The performance of Text-to-SQL has been greatly enhanced with the emergence of Large Language Models (LLMs). In this survey, we provide a comprehensive review of Text-to-SQL techniques powered by LLMs, covering its entire lifecycle from the following four aspects: (1) Model: Text-to-SQL translation techniques that tackle not only NL ambiguity and under-specification, but also properly map NL with database schema and instances; (2) Data: From the collection of training data, data synthesis due to training data scarcity, to Text-to-SQL benchmarks; (3) Evaluation: Evaluating Text-to-SQL methods from multiple angles using different metrics and granularities; and (4) Error Analysis: analyzing Text-to-SQL errors to find the root cause and guiding Text-to-SQL models to evolve. Moreover, we offer a rule of thumb for developing Text-to-SQL solutions. Finally, we discuss the research challenges and open problems of Text-to-SQL in the LLMs era.
Authors: Guhong Chen, Liyang Fan, Zihan Gong, Nan Xie, Zixuan Li, Ziqiang Liu, Chengming Li, Qiang Qu, Hamid Alinejad-Rokny, Shiwen Ni, Min Yang
Abstract: Current research in LLM-based simulation systems lacks comprehensive solutions for modeling real-world court proceedings, while existing legal language models struggle with dynamic courtroom interactions. We present AgentCourt, a comprehensive legal simulation framework that addresses these challenges through adversarial evolution of LLM-based agents. Our AgentCourt introduces a new adversarial evolutionary approach for agents called AdvEvol, which performs dynamic knowledge learning and evolution through structured adversarial interactions in a simulated courtroom program, breaking the limitations of the traditional reliance on static knowledge bases or manual annotations. By simulating 1,000 civil cases, we construct an evolving knowledge base that enhances the agents' legal reasoning abilities. The evolved lawyer agents demonstrated outstanding performance on our newly introduced CourtBench benchmark, achieving a 12.1% improvement in performance compared to the original lawyer agents. Evaluations by professional lawyers confirm the effectiveness of our approach across three critical dimensions: cognitive agility, professional knowledge, and logical rigor. Beyond outperforming specialized legal models in interactive reasoning tasks, our findings emphasize the importance of adversarial learning in legal AI and suggest promising directions for extending simulation-based legal reasoning to broader judicial and regulatory contexts. The project's code is available at: https://github.com/relic-yuexi/AgentCourt
Authors: Lin Zuo, Yongqi Ding, Mengmeng Jing, Kunshan Yang, Biao Chen, Yunqian Yu
Abstract: This paper explores the application of spiking neural networks (SNNs), known for their low-power binary spikes, to bearing fault diagnosis, bridging the gap between high-performance AI algorithms and real-world industrial scenarios. In particular, we identify two key limitations of existing SNN fault diagnosis methods: inadequate encoding capacity that necessitates cumbersome data preprocessing, and non-spike-oriented architectures that constrain the performance of SNNs. To alleviate these problems, we propose a Multi-scale Residual Attention SNN (MRA-SNN) to simultaneously improve the efficiency, performance, and robustness of SNN methods. By incorporating a lightweight attention mechanism, we have designed a multi-scale attention encoding module to extract multiscale fault features from vibration signals and encode them as spatio-temporal spikes, eliminating the need for complicated preprocessing. Then, the spike residual attention block extracts high-dimensional fault features and enhances the expressiveness of sparse spikes with the attention mechanism for end-to-end diagnosis. In addition, the performance and robustness of MRA-SNN is further enhanced by introducing the lightweight attention mechanism within the spiking neurons to simulate the biological dendritic filtering effect. Extensive experiments on MFPT, JNU, Bearing, and Gearbox benchmark datasets demonstrate that MRA-SNN significantly outperforms existing methods in terms of accuracy, energy consumption, and noise robustness, and is more feasible for deployment in real-world industrial scenarios.
Authors: Yizhan Li, Sifan Wu, Christopher Smith, Thomas Lo, Bang Liu
Abstract: Writing clinical notes and documenting medical exams is a critical task for healthcare professionals, serving as a vital component of patient care documentation. However, manually writing these notes is time-consuming and can impact the amount of time clinicians can spend on direct patient interaction and other tasks. Consequently, the development of automated clinical note generation systems has emerged as a clinically meaningful area of research within AI for health. In this paper, we present three key contributions to the field of clinical note generation using large language models (LLMs). First, we introduce CliniKnote, a comprehensive dataset consisting of 1,200 complex doctor-patient conversations paired with their full clinical notes. This dataset, created and curated by medical experts with the help of modern neural networks, provides a valuable resource for training and evaluating models in clinical note generation tasks. Second, we propose the K-SOAP (Keyword, Subjective, Objective, Assessment, and Plan) note format, which enhances traditional SOAP~\cite{podder2023soap} (Subjective, Objective, Assessment, and Plan) notes by adding a keyword section at the top, allowing for quick identification of essential information. Third, we develop an automatic pipeline to generate K-SOAP notes from doctor-patient conversations and benchmark various modern LLMs using various metrics. Our results demonstrate significant improvements in efficiency and performance compared to standard LLM finetuning methods.
Authors: Jo\~ao Figueiredo, Afonso Carvalho, Daniel Castro, Daniel Gon\c{c}alves, Nuno Santos
Abstract: A vishing attack is a form of social engineering where attackers use phone calls to deceive individuals into disclosing sensitive information, such as personal data, financial information, or security credentials. Attackers exploit the perceived urgency and authenticity of voice communication to manipulate victims, often posing as legitimate entities like banks or tech support. Vishing is a particularly serious threat as it bypasses security controls designed to protect information. In this work, we study the potential for vishing attacks to escalate with the advent of AI. In theory, AI-powered software bots may have the ability to automate these attacks by initiating conversations with potential victims via phone calls and deceiving them into disclosing sensitive information. To validate this thesis, we introduce ViKing, an AI-powered vishing system developed using publicly available AI technology. It relies on a Large Language Model (LLM) as its core cognitive processor to steer conversations with victims, complemented by a pipeline of speech-to-text and text-to-speech modules that facilitate audio-text conversion in phone calls. Through a controlled social experiment involving 240 participants, we discovered that ViKing has successfully persuaded many participants to reveal sensitive information, even those who had been explicitly warned about the risk of vishing campaigns. Interactions with ViKing's bots were generally considered realistic. From these findings, we conclude that tools like ViKing may already be accessible to potential malicious actors, while also serving as an invaluable resource for cyber awareness programs.
Authors: Sebastian Bordt, Suraj Srinivas, Valentyn Boreiko, Ulrike von Luxburg
Abstract: The leakage of benchmark data into the training data has emerged as a significant challenge for evaluating the capabilities of large language models (LLMs). In this work, we challenge the common assumption that small-scale contamination renders benchmark evaluations invalid. First, we experimentally quantify the magnitude of benchmark overfitting based on scaling along three dimensions: The number of model parameters (up to 1.6B), the number of times an example is seen (up to 144), and the number of training tokens (up to 40B). If model and data follow the Chinchilla scaling laws, minor contamination indeed leads to overfitting. At the same time, even 144 times of contamination can be forgotten if the training data is scaled beyond five times Chinchilla, a regime characteristic of many modern LLMs. Continual pre-training of OLMo-7B corroborates these results. Next, we study the impact of the weight decay parameter on example forgetting, showing that empirical forgetting occurs faster than the cumulative weight decay. This allows us to gauge the degree of example forgetting in large-scale training runs, indicating that many LLMs, including Lllama 3 405B, have forgotten the data seen at the beginning of training.
Authors: Zian Li, Cai Zhou, Xiyuan Wang, Xingang Peng, Muhan Zhang
Abstract: Recent advances in molecular generative models have demonstrated great promise for accelerating scientific discovery, particularly in drug design. However, these models often struggle to generate high-quality molecules, especially in conditional scenarios where specific molecular properties must be satisfied. In this work, we introduce GeoRCG, a general framework to improve molecular generative models by integrating geometric representation conditions with provable theoretical guarantees. We decompose the generation process into two stages: first, generating an informative geometric representation; second, generating a molecule conditioned on the representation. Compared with single-stage generation, the easy-to-generate representation in the first stage guides the second stage generation toward a high-quality molecule in a goal-oriented way. Leveraging EDM and SemlaFlow as base generators, we observe significant quality improvements in unconditional molecule generation on the widely used QM9 and GEOM-DRUG datasets. More notably, in the challenging conditional molecular generation task, our framework achieves an average 50\% performance improvement over state-of-the-art approaches, highlighting the superiority of conditioning on semantically rich geometric representations. Furthermore, with such representation guidance, the number of diffusion steps can be reduced to as small as 100 while largely preserving the generation quality achieved with 1,000 steps, thereby significantly reducing the generation iterations needed.
Authors: Ethan He, Abhinav Khattar, Ryan Prenger, Vijay Korthikanti, Zijie Yan, Tong Liu, Shiqing Fan, Ashwath Aithal, Mohammad Shoeybi, Bryan Catanzaro
Abstract: Upcycling pre-trained dense language models into sparse mixture-of-experts (MoE) models is an efficient approach to increase the model capacity of already trained models. However, optimal techniques for upcycling at scale remain unclear. In this work, we conduct an extensive study of upcycling methods and hyperparameters for billion-parameter scale language models. We propose a novel "virtual group" initialization scheme and weight scaling approach to enable upcycling into fine-grained MoE architectures. Through ablations, we find that upcycling outperforms continued dense model training. In addition, we show that softmax-then-topK expert routing improves over topK-then-softmax approach and higher granularity MoEs can help improve accuracy. Finally, we upcycled Nemotron-4 15B on 1T tokens and compared it to a continuously trained version of the same model on the same 1T tokens: the continuous trained model achieved 65.3% MMLU, whereas the upcycled model achieved 67.6%. Our results offer insights and best practices to effectively leverage upcycling for building MoE language models. Code is available.
Authors: Amit Kumar Singh, Vrijendra Singh
Abstract: Deep learning and advancements in contactless sensors have significantly enhanced our ability to understand complex human activities in healthcare settings. In particular, deep learning models utilizing computer vision have been developed to enable detailed analysis of human gesture recognition, especially repetitive gestures which are commonly observed behaviors in children with autism. This research work aims to identify repetitive behaviors indicative of autism by analyzing videos captured in natural settings as children engage in daily activities. The focus is on accurately categorizing real-time repetitive gestures such as spinning, head banging, and arm flapping. To this end, we utilize the publicly accessible Self-Stimulatory Behavior Dataset (SSBD) to classify these stereotypical movements. A key component of the proposed methodology is the use of \textbf{VideoMAE}, a model designed to improve both spatial and temporal analysis of video data through a masking and reconstruction mechanism. This model significantly outperformed traditional methods, achieving an accuracy of 97.7\%, a 14.7\% improvement over the previous state-of-the-art.
Authors: Asger Horn Brorholt, Kim Guldstrand Larsen, Christian Schilling
Abstract: Deep reinforcement learning has emerged as a powerful tool for obtaining high-performance policies. However, the safety of these policies has been a long-standing issue. One promising paradigm to guarantee safety is a shield, which shields a policy from making unsafe actions. However, computing a shield scales exponentially in the number of state variables. This is a particular concern in multi-agent systems with many agents. In this work, we propose a novel approach for multi-agent shielding. We address scalability by computing individual shields for each agent. The challenge is that typical safety specifications are global properties, but the shields of individual agents only ensure local properties. Our key to overcome this challenge is to apply assume-guarantee reasoning. Specifically, we present a sound proof rule that decomposes a (global, complex) safety specification into (local, simple) obligations for the shields of the individual agents. Moreover, we show that applying the shields during reinforcement learning significantly improves the quality of the policies obtained for a given training budget. We demonstrate the effectiveness and scalability of our multi-agent shielding framework in two case studies, reducing the computation time from hours to seconds and achieving fast learning convergence.
Authors: Kajetan Schweighofer, Lukas Aichberger, Mykyta Ielanskyi, Sepp Hochreiter
Abstract: Reliable estimation of predictive uncertainty is crucial for machine learning applications, particularly in high-stakes scenarios where hedging against risks is essential. Despite its significance, there is no universal agreement on how to best quantify predictive uncertainty. In this work, we revisit core concepts to propose a framework for information-theoretic measures of predictive uncertainty. Our proposed framework categorizes predictive uncertainty measures according to two factors: (I) The predicting model (II) The approximation of the true predictive distribution. Examining all possible combinations of these two factors, we derive a set of predictive uncertainty measures that includes both known and newly introduced ones. We extensively evaluate these measures across a broad set of tasks, identifying conditions under which certain measures excel. Our findings show the importance of aligning the choice of uncertainty measure with the predicting model on in-distribution (ID) data, the limitations of epistemic uncertainty measures for out-of-distribution (OOD) data, and that the disentanglement between measures varies substantially between ID and OOD data. Together, these insights provide a more comprehensive understanding of predictive uncertainty measures, revealing their implicit assumptions and relationships.
Authors: Peng Jin, Bo Zhu, Li Yuan, Shuicheng Yan
Abstract: In this work, we upgrade the multi-head attention mechanism, the core of the Transformer model, to improve efficiency while maintaining or surpassing the previous accuracy level. We show that multi-head attention can be expressed in the summation form. Drawing on the insight that not all attention heads hold equal significance, we propose Mixture-of-Head attention (MoH), a new architecture that treats attention heads as experts in the Mixture-of-Experts (MoE) mechanism. MoH has two significant advantages: First, MoH enables each token to select the appropriate attention heads, enhancing inference efficiency without compromising accuracy or increasing the number of parameters. Second, MoH replaces the standard summation in multi-head attention with a weighted summation, introducing flexibility to the attention mechanism and unlocking extra performance potential. Extensive experiments on ViT, DiT, and LLMs demonstrate that MoH outperforms multi-head attention by using only 50%-90% of the attention heads. Moreover, we demonstrate that pre-trained multi-head attention models, such as LLaMA3-8B, can be further continue-tuned into our MoH models. Notably, MoH-LLaMA3-8B achieves an average accuracy of 64.0% across 14 benchmarks, outperforming LLaMA3-8B by 2.4% by utilizing only 75% of the attention heads. We believe the proposed MoH is a promising alternative to multi-head attention and provides a strong foundation for developing advanced and efficient attention-based models.
Authors: Yuxiang Wang, Jianzhong Qi, Junhao Gan
Abstract: Question answering on free-form tables (a.k.a. TableQA) is a challenging task because of the flexible structure and complex schema of tables. Recent studies use Large Language Models (LLMs) for this task, exploiting their capability in understanding the questions and tabular data, which are typically given in natural language and contain many textual fields, respectively. While this approach has shown promising results, it overlooks the challenges brought by numerical values which are common in tabular data, and LLMs are known to struggle with such values. We aim to address this issue, and we propose a model named TabLaP that uses LLMs as a planner rather than an answer generator. This approach exploits LLMs' capability in multi-step reasoning while leaving the actual numerical calculations to a Python interpreter for accurate calculation. Recognizing the inaccurate nature of LLMs, we further make a first attempt to quantify the trustworthiness of the answers produced by TabLaP, such that users can use TabLaP in a regret-aware manner. Experimental results on two benchmark datasets show that TabLaP is substantially more accurate than the state-of-the-art models, improving the answer accuracy by 5.7% and 5.8% on the two datasets, respectively.
Authors: Yuji Wang, Zehua Chen, Xiaoyu Chen, Yixiang Wei, Jun Zhu, Jianfei Chen
Abstract: Diffusion models have achieved remarkable progress on image-to-video (I2V) generation, while their noise-to-data generation process is inherently mismatched with this task, which may lead to suboptimal synthesis quality. In this work, we present FrameBridge. By modeling the frame-to-frames generation process with a bridge model based data-to-data generative process, we are able to fully exploit the information contained in the given image and improve the consistency between the generation process and I2V task. Moreover, we propose two novel techniques toward the two popular settings of training I2V models, respectively. Firstly, we propose SNR-Aligned Fine-tuning (SAF), making the first attempt to fine-tune a diffusion model to a bridge model and, therefore, allowing us to utilize the pre-trained diffusion-based text-to-video (T2V) models. Secondly, we propose neural prior, further improving the synthesis quality of FrameBridge when training from scratch. Experiments conducted on WebVid-2M and UCF-101 demonstrate the superior quality of FrameBridge in comparison with the diffusion counterpart (zero-shot FVD 95 vs. 192 on MSR-VTT and non-zero-shot FVD 122 vs. 171 on UCF-101), and the advantages of our proposed SAF and neural prior for bridge-based I2V models. The project page: https://framebridge-icml.github.io/.
Authors: Ian Gemp, Andreas Haupt, Luke Marris, Siqi Liu, Georgios Piliouras
Abstract: Behavioral diversity, expert imitation, fairness, safety goals and others give rise to preferences in sequential decision making domains that do not decompose additively across time. We introduce the class of convex Markov games that allow general convex preferences over occupancy measures. Despite infinite time horizon and strictly higher generality than Markov games, pure strategy Nash equilibria exist. Furthermore, equilibria can be approximated empirically by performing gradient descent on an upper bound of exploitability. Our experiments reveal novel solutions to classic repeated normal-form games, find fair solutions in a repeated asymmetric coordination game, and prioritize safe long-term behavior in a robot warehouse environment. In the prisoner's dilemma, our algorithm leverages transient imitation to find a policy profile that deviates from observed human play only slightly, yet achieves higher per-player utility while also being three orders of magnitude less exploitable.
Authors: Iman Saberi, Fatemeh Fard
Abstract: Large Language Models (LLMs) and Code-LLMs (CLLMs) have significantly improved code generation, but, they frequently face difficulties when dealing with challenging and complex problems. Retrieval-Augmented Generation (RAG) addresses this issue by retrieving and integrating external knowledge at the inference time. However, retrieval models often fail to find most relevant context, and generation models, with limited context capacity, can hallucinate when given irrelevant data. We present a novel framework that leverages a Programming Knowledge Graph (PKG) to semantically represent and retrieve code. This approach enables fine-grained code retrieval by focusing on the most relevant segments while reducing irrelevant context through a tree-pruning technique. PKG is coupled with a re-ranking mechanism to reduce even more hallucinations by selectively integrating non-RAG solutions. We propose two retrieval approaches-block-wise and function-wise-based on the PKG, optimizing context granularity. Evaluations on the HumanEval and MBPP benchmarks show our method improves pass@1 accuracy by up to 20%, and outperforms state-of-the-art models by up to 34% on MBPP. Our contributions include PKG-based retrieval, tree pruning to enhance retrieval precision, a re-ranking method for robust solution selection and a Fill-in-the-Middle (FIM) enhancer module for automatic code augmentation with relevant comments and docstrings.
Authors: Yifan Deng, Spencer S. Ericksen, Anthony Gitter
Abstract: The development of large language models and multi-modal models has enabled the appealing idea of generating novel molecules from text descriptions. Generative modeling would shift the paradigm from relying on large-scale chemical screening to find molecules with desired properties to directly generating those molecules. However, multi-modal models combining text and molecules are often trained from scratch, without leveraging existing high-quality pretrained models. Training from scratch consumes more computational resources and prohibits model scaling. In contrast, we propose a lightweight adapter-based strategy named Chemical Language Model Linker (ChemLML). ChemLML blends the two single domain models and obtains conditional molecular generation from text descriptions while still operating in the specialized embedding spaces of the molecular domain. ChemLML can tailor diverse pretrained text models for molecule generation by training relatively few adapter parameters. We find that the choice of molecular representation used within ChemLML, SMILES versus SELFIES, has a strong influence on conditional molecular generation performance. SMILES is often preferable despite not guaranteeing valid molecules. We raise issues in using the entire PubChem dataset of molecules and their associated descriptions for evaluating molecule generation and provide a filtered version of the dataset as a generation test set. To demonstrate how ChemLML could be used in practice, we generate candidate protein inhibitors and use docking to assess their quality and also generate candidate membrane permeable molecules.
Authors: Shuchen Wu, Mirko Thalmann, Peter Dayan, Zeynep Akata, Eric Schulz
Abstract: Humans excel at learning abstract patterns across different sequences, filtering out irrelevant details, and transferring these generalized concepts to new sequences. In contrast, many sequence learning models lack the ability to abstract, which leads to memory inefficiency and poor transfer. We introduce a non-parametric hierarchical variable learning model (HVM) that learns chunks from sequences and abstracts contextually similar chunks as variables. HVM efficiently organizes memory while uncovering abstractions, leading to compact sequence representations. When learning on language datasets such as babyLM, HVM learns a more efficient dictionary than standard compression algorithms such as Lempel-Ziv. In a sequence recall task requiring the acquisition and transfer of variables embedded in sequences, we demonstrate HVM's sequence likelihood correlates with human recall times. In contrast, large language models (LLMs) struggle to transfer abstract variables as effectively as humans. From HVM's adjustable layer of abstraction, we demonstrate that the model realizes a precise trade-off between compression and generalization. Our work offers a cognitive model that captures the learning and transfer of abstract representations in human cognition and differentiates itself from LLMs.
Authors: Ryan Liu, Jiayi Geng, Addison J. Wu, Ilia Sucholutsky, Tania Lombrozo, Thomas L. Griffiths
Abstract: Chain-of-thought (CoT) prompting has become a widely used strategy for improving large language and multimodal model performance. However, it is still an open question under which settings CoT systematically reduces performance. In this paper, we seek to identify the characteristics of tasks where CoT reduces performance by drawing inspiration from cognitive psychology, focusing on six representative tasks from the psychological literature where deliberation hurts performance in humans. In three of these tasks, state-of-the-art models exhibit significant performance drop-offs with CoT (up to 36.3\% absolute accuracy for OpenAI o1-preview compared to GPT-4o), while in others, CoT effects are mixed, with positive, neutral, and negative changes. While models and humans do not exhibit perfectly parallel cognitive processes, considering cases where thinking has negative consequences for humans helps identify settings where it negatively impacts models. By connecting the literature on human verbal thinking and deliberation with evaluations of CoT, we offer a perspective for understanding the impact of inference-time reasoning.
Authors: Shuai Niu, Jing Ma, Hongzhan Lin, Liang Bai, Zhihua Wang, Yida Xu, Yunya Song, Xian Yang
Abstract: Interpretation is critical for disease diagnosis, but existing models struggle to balance predictive accuracy with human-understandable rationales. While large language models (LLMs) offer strong reasoning abilities, their clinical use is limited by high computational costs and restricted multimodal reasoning ability. Small language models (SLMs) are efficient but lack advanced reasoning for integrating multimodal medical data. In addition, both LLMs and SLMs lack domain knowledge for trustworthy reasoning. Therefore, we propose ClinRaGen, enhancing SLMs by leveraging LLM-derived reasoning ability via rationale distillation and domain knowledge injection for trustworthy multimodal rationale generation. Key innovations include a sequential rationale distillation framework that equips SLMs with LLM-comparable multimodal reasoning abilities, and a knowledge-augmented attention mechanism that jointly unifies multimodal representation from time series and textual data in the same encoding space, enabling it to be naturally interpreted by SLMs while incorporating domain knowledge for reliable rationale generation. Experiments on real-world medical datasets show that ClinRaGen achieves state-of-the-art performance in disease diagnosis and rationale generation, demonstrating the effectiveness of combining LLM-driven reasoning with knowledge augmentation for improved interpretability.
Authors: Caspar Oesterheld, Emery Cooper, Miles Kodama, Linh Chi Nguyen, Ethan Perez
Abstract: We introduce a dataset of natural-language questions in the decision theory of so-called Newcomb-like problems. Newcomb-like problems include, for instance, decision problems in which an agent interacts with a similar other agent, and thus has to reason about the fact that the other agent will likely reason in similar ways. Evaluating LLM reasoning about Newcomb-like problems is important because interactions between foundation-model-based agents will often be Newcomb-like. Some ways of reasoning about Newcomb-like problems may allow for greater cooperation between models. Our dataset contains both capabilities questions (i.e., questions with a unique, uncontroversially correct answer) and attitude questions (i.e., questions about which decision theorists would disagree). We use our dataset for an investigation of decision-theoretical capabilities and expressed attitudes and their interplay in existing models (different models by OpenAI, Anthropic, Meta, GDM, Reka, etc.), as well as models under simple prompt-based interventions. We find, among other things, that attitudes vary significantly between existing models; that high capabilities are associated with attitudes more favorable toward so-called evidential decision theory; and that attitudes are consistent across different types of questions.
Authors: Amar Abane, Anis Bekri, Abdella Battou, Saddek Bensalem
Abstract: Efficiently processing and interpreting network data is critical for the operation of increasingly complex networks. Recent advances in Large Language Models (LLM) and Retrieval-Augmented Generation (RAG) techniques have improved data processing in network management. However, existing RAG methods like VectorRAG and GraphRAG struggle with the complexity and implicit nature of semi-structured technical data, leading to inefficiencies in time, cost, and retrieval. This paper introduces FastRAG, a novel RAG approach designed for semi-structured data. FastRAG employs schema learning and script learning to extract and structure data without needing to submit entire data sources to an LLM. It integrates text search with knowledge graph (KG) querying to improve accuracy in retrieving context-rich information. Evaluation results demonstrate that FastRAG provides accurate question answering, while improving up to 90% in time and 85% in cost compared to GraphRAG.
Authors: Soumya Suvra Ghosal, Souradip Chakraborty, Vaibhav Singh, Tianrui Guan, Mengdi Wang, Alvaro Velasquez, Ahmad Beirami, Furong Huang, Dinesh Manocha, Amrit Singh Bedi
Abstract: With the widespread deployment of Multimodal Large Language Models (MLLMs) for visual-reasoning tasks, improving their safety has become crucial. Recent research indicates that despite training-time safety alignment, these models remain vulnerable to jailbreak attacks. In this work, we first highlight an important safety gap to describe that alignment achieved solely through safety training may be insufficient against jailbreak attacks. To address this vulnerability, we propose Immune, an inference-time defense framework that leverages a safe reward model through controlled decoding to defend against jailbreak attacks. Additionally, we provide a mathematical characterization of Immune, offering insights on why it improves safety against jailbreaks. Extensive evaluations on diverse jailbreak benchmarks using recent MLLMs reveal that Immune effectively enhances model safety while preserving the model's original capabilities. For instance, against text-based jailbreak attacks on LLaVA-1.6, Immune reduces the attack success rate by 57.82% and 16.78% compared to the base MLLM and state-of-the-art defense strategy, respectively.
Authors: Zhihang Lin, Mingbao Lin, Wengyi Zhan, Rongrong Ji
Abstract: Diffusion models suffer severe object repetition and local distortion when the inference resolution differs from its pre-trained resolution. We propose AccDiffusion v2, an accurate method for patch-wise higher-resolution diffusion extrapolation without training. Our in-depth analysis in this paper shows that using an identical text prompt for different patches leads to repetitive generation, while the absence of a prompt undermines image details. In response, our AccDiffusion v2 novelly decouples the vanilla image-content-aware prompt into a set of patch-content-aware prompts, each of which serves as a more precise description of a patch. Further analysis reveals that local distortion arises from inaccurate descriptions in prompts about the local structure of higher-resolution images. To address this issue, AccDiffusion v2, for the first time, introduces an auxiliary local structural information through ControlNet during higher-resolution diffusion extrapolation aiming to mitigate the local distortions. Finally, our analysis indicates that global semantic information is conducive to suppressing both repetitive generation and local distortion. Hence, our AccDiffusion v2 further proposes dilated sampling with window interaction for better global semantic information during higher-resolution diffusion extrapolation. We conduct extensive experiments, including both quantitative and qualitative comparisons, to demonstrate the efficacy of our AccDiffusion v2. The quantitative comparison shows that AccDiffusion v2 achieves state-of-the-art performance in image generation extrapolation without training. The qualitative comparison intuitively illustrates that AccDiffusion v2 effectively suppresses the issues of repetitive generation and local distortion in image generation extrapolation. Our code is available at https://github.com/lzhxmu/AccDiffusion_v2.
Authors: Guangda Liu, Chengwei Li, Jieru Zhao, Chenqi Zhang, Minyi Guo
Abstract: Large Language Models (LLMs) have been widely deployed in a variety of applications, and the context length is rapidly increasing to handle tasks such as long-document QA and complex logical reasoning. However, long context poses significant challenges for inference efficiency, including high memory costs of key-value (KV) cache and increased latency due to extensive memory accesses. Recent works have proposed compressing KV cache to approximate computation, but these methods either evict tokens permanently, never recalling them for later inference, or recall previous tokens at the granularity of pages divided by textual positions. Both approaches degrade the model accuracy and output quality. To achieve efficient and accurate recallable KV cache compression, we introduce ClusterKV, which recalls tokens at the granularity of semantic clusters. We design and implement efficient algorithms and systems for clustering, selection, indexing and caching. Experiment results show that ClusterKV attains negligible accuracy loss across various tasks with 32k context lengths, using only a 1k to 2k KV cache budget, and achieves up to a 2$\times$ speedup in latency and a 2.5$\times$ improvement in decoding throughput. Compared to SoTA recallable KV compression methods, ClusterKV demonstrates higher model accuracy and output quality, while maintaining or exceeding inference efficiency. Our code is available at https://github.com/sjtu-zhao-lab/ClusterKV.
Authors: Yue Jin, Shuangqing Wei, Giovanni Montana
Abstract: In human society, the conflict between self-interest and collective well-being often obstructs efforts to achieve shared welfare. Related concepts like the Tragedy of the Commons and Social Dilemmas frequently manifest in our daily lives. As artificial agents increasingly serve as autonomous proxies for humans, we propose a novel multi-agent reinforcement learning (MARL) method to address this issue - learning policies to maximise collective returns even when individual agents' interests conflict with the collective one. Unlike traditional cooperative MARL solutions that involve sharing rewards, values, and policies or designing intrinsic rewards to encourage agents to learn collectively optimal policies, we propose a novel MARL approach where agents exchange action suggestions. Our method reveals less private information compared to sharing rewards, values, or policies, while enabling effective cooperation without the need to design intrinsic rewards. Our algorithm is supported by our theoretical analysis that establishes a bound on the discrepancy between collective and individual objectives, demonstrating how sharing suggestions can align agents' behaviours with the collective objective. Experimental results demonstrate that our algorithm performs competitively with baselines that rely on value or policy sharing or intrinsic rewards.
Authors: Haoyi Zhang, Shizhao Sun, Yibo Lin, Runsheng Wang, Jiang Bian
Abstract: Analog circuits are crucial in modern electronic systems, and automating their design has attracted significant research interest. One of major challenges is topology synthesis, which determines circuit components and their connections. Recent studies explore large language models (LLM) for topology synthesis. However, the scenarios addressed by these studies do not align well with practical applications. Specifically, existing work uses vague design requirements as input and outputs an ideal model, but detailed structural requirements and device-level models are more practical. Moreover, current approaches either formulate topology synthesis as graph generation or Python code generation, whereas practical topology design is a complex process that demands extensive design knowledge. In this work, we propose AnalogXpert, a LLM-based agent aiming at solving practical topology synthesis problem by incorporating circuit design expertise into LLMs. First, we represent analog topology as SPICE code and introduce a subcircuit library to reduce the design space, in the same manner as experienced designers. Second, we decompose the problem into two sub-task (i.e., block selection and block connection) through the use of CoT and incontext learning techniques, to mimic the practical design process. Third, we introduce a proofreading strategy that allows LLMs to incrementally correct the errors in the initial design, akin to human designers who iteratively check and adjust the initial topology design to ensure accuracy. Finally, we construct a high-quality benchmark containing both real data (30) and synthetic data (2k). AnalogXpert achieves 40% and 23% success rates on the synthetic dataset and real dataset respectively, which is markedly better than those of GPT-4o (3% on both the synthetic dataset and the real dataset).
Authors: Chuanzhi Xu, Langyi Chen, Haodong Chen, Vera Chung, Qiang Qu
Abstract: Neuromorphic cameras, also known as event cameras, are asynchronous brightness-change sensors that can capture extremely fast motion without suffering from motion blur, making them particularly promising for 3D reconstruction in extreme environments. However, existing research on 3D reconstruction using monocular neuromorphic cameras is limited, and most of the methods rely on estimating physical priors and employ complex multi-step pipelines. In this work, we propose an end-to-end method for dense voxel 3D reconstruction using neuromorphic cameras that eliminates the need to estimate physical priors. Our method incorporates a novel event representation to enhance edge features, enabling the proposed feature-enhancement model to learn more effectively. Additionally, we introduced Optimal Binarization Threshold Selection Principle as a guideline for future related work, using the optimal reconstruction results achieved with threshold optimization as the benchmark. Our method achieves a 54.6% improvement in reconstruction accuracy compared to the baseline method.
Authors: Jian Lang, Zhangtao Cheng, Ting Zhong, Fan Zhou
Abstract: Multimodal learning with incomplete modality is practical and challenging. Recently, researchers have focused on enhancing the robustness of pre-trained MultiModal Transformers (MMTs) under missing modality conditions by applying learnable prompts. However, these prompt-based methods face several limitations: (1) incomplete modalities provide restricted modal cues for task-specific inference, (2) dummy imputation for missing content causes information loss and introduces noise, and (3) static prompts are instance-agnostic, offering limited knowledge for instances with various missing conditions. To address these issues, we propose RAGPT, a novel Retrieval-AuGmented dynamic Prompt Tuning framework. RAGPT comprises three modules: (I) the multi-channel retriever, which identifies similar instances through a within-modality retrieval strategy, (II) the missing modality generator, which recovers missing information using retrieved contexts, and (III) the context-aware prompter, which captures contextual knowledge from relevant instances and generates dynamic prompts to largely enhance the MMT's robustness. Extensive experiments conducted on three real-world datasets show that RAGPT consistently outperforms all competitive baselines in handling incomplete modality problems. The code of our work and prompt-based baselines is available at https://github.com/Jian-Lang/RAGPT.
Authors: Fan Bu, Zheng Wang, Siyi Wang, Ziyao Liu
Abstract: As Large Language Models (LLMs) become increasingly prevalent in tasks related to cultural heritage, such as generating descriptions of historical monuments, translating ancient texts, preserving oral traditions, and creating educational content, their ability to produce accurate and culturally aligned texts is being increasingly relied upon by users and researchers. However, cultural value misalignments may exist in generated texts, such as the misrepresentation of historical facts, the erosion of cultural identity, and the oversimplification of complex cultural narratives, which may lead to severe consequences. Therefore, investigating value misalignment in the context of LLM for cultural heritage is crucial for mitigating these risks, yet there has been a significant lack of systematic and comprehensive study and investigation in this area. To fill this gap, we systematically assess the reliability of LLMs in generating culturally aligned texts for cultural heritage-related tasks. We conduct a comprehensive evaluation by compiling an extensive set of 1066 query tasks covering 5 widely recognized categories with 17 aspects within the knowledge framework of cultural heritage across 5 open-source LLMs, and examine both the type and rate of cultural value misalignments in the generated texts. Using both automated and manual approaches, we effectively detect and analyze the cultural value misalignments in LLM-generated texts. Our findings are concerning: over 65% of the generated texts exhibit notable cultural misalignments, with certain tasks demonstrating almost complete misalignment with key cultural values. Beyond these findings, this paper introduces a benchmark dataset and a comprehensive evaluation workflow that can serve as a valuable resource for future research aimed at enhancing the cultural sensitivity and reliability of LLMs.
Authors: Shucong Zhang, Titouan Parcollet, Rogier van Dalen, Sourav Bhattacharya
Abstract: Self-attention relies on positional embeddings to encode input order. Relative Position (RelPos) embeddings are widely used in Automatic Speech Recognition (ASR). However, RelPos has quadratic time complexity to input length and is often incompatible with fast GPU implementations of attention. In contrast, Rotary Positional Embedding (RoPE) rotates each input vector based on its absolute position, taking linear time to sequence length, implicitly encoding relative distances through self-attention dot products. Thus, it is usually compatible with efficient attention. However, its use in ASR remains underexplored. This work evaluates RoPE across diverse ASR tasks with training data ranging from 100 to 50,000 hours, covering various speech types (read, spontaneous, clean, noisy) and different accents in both streaming and non-streaming settings. ASR error rates are similar or better than RelPos, while training time is reduced by up to 21%. Code is available via the SpeechBrain toolkit.
Authors: Tong Xiao, Jingbo Zhu
Abstract: This is a book about large language models. As indicated by the title, it primarily focuses on foundational concepts rather than comprehensive coverage of all cutting-edge technologies. The book is structured into five main chapters, each exploring a key area: pre-training, generative models, prompting, alignment, and inference. It is intended for college students, professionals, and practitioners in natural language processing and related fields, and can serve as a reference for anyone interested in large language models.
Authors: Sili Chen, Hengkai Guo, Shengnan Zhu, Feihu Zhang, Zilong Huang, Jiashi Feng, Bingyi Kang
Abstract: Depth Anything has achieved remarkable success in monocular depth estimation with strong generalization ability. However, it suffers from temporal inconsistency in videos, hindering its practical applications. Various methods have been proposed to alleviate this issue by leveraging video generation models or introducing priors from optical flow and camera poses. Nonetheless, these methods are only applicable to short videos (< 10 seconds) and require a trade-off between quality and computational efficiency. We propose Video Depth Anything for high-quality, consistent depth estimation in super-long videos (over several minutes) without sacrificing efficiency. We base our model on Depth Anything V2 and replace its head with an efficient spatial-temporal head. We design a straightforward yet effective temporal consistency loss by constraining the temporal depth gradient, eliminating the need for additional geometric priors. The model is trained on a joint dataset of video depth and unlabeled images, similar to Depth Anything V2. Moreover, a novel key-frame-based strategy is developed for long video inference. Experiments show that our model can be applied to arbitrarily long videos without compromising quality, consistency, or generalization ability. Comprehensive evaluations on multiple video benchmarks demonstrate that our approach sets a new state-of-the-art in zero-shot video depth estimation. We offer models of different scales to support a range of scenarios, with our smallest model capable of real-time performance at 30 FPS.
Authors: Yan Chen, Qinxun Bai, Yiteng Zhang, Shi Dong, Maria Dimakopoulou, Qi Sun, Zhengyuan Zhou
Abstract: Designing learning agents that explore efficiently in a complex environment has been widely recognized as a fundamental challenge in reinforcement learning. While a number of works have demonstrated the effectiveness of techniques based on randomized value functions on a single agent, it remains unclear, from a theoretical point of view, whether injecting randomization can help a society of agents {\it concurently} explore an environment. The theoretical results %that we established in this work tender an affirmative answer to this question. We adapt the concurrent learning framework to \textit{randomized least-squares value iteration} (RLSVI) with \textit{aggregated state representation}. We demonstrate polynomial worst-case regret bounds in both finite- and infinite-horizon environments. In both setups the per-agent regret decreases at an optimal rate of $\Theta\left(\frac{1}{\sqrt{N}}\right)$, highlighting the advantage of concurent learning. Our algorithm exhibits significantly lower space complexity compared to \cite{russo2019worst} and \cite{agrawal2021improved}. We reduce the space complexity by a factor of $K$ while incurring only a $\sqrt{K}$ increase in the worst-case regret bound, compared to \citep{agrawal2021improved,russo2019worst}. Additionally, we conduct numerical experiments to demonstrate our theoretical findings.
Authors: Naihao Deng, Rada Mihalcea
Abstract: Recent advances in table understanding have focused on instruction-tuning large language models (LLMs) for table-related tasks. However, existing research has overlooked the impact of hyperparameter choices, and also lacks a comprehensive evaluation of the out-of-domain table understanding ability and the general capabilities of these table LLMs. In this paper, we evaluate these abilities in existing table LLMs, and find significant declines in both out-of-domain table understanding and general capabilities as compared to their base models. Through systematic analysis, we show that hyperparameters, such as learning rate, can significantly influence both table-specific and general capabilities. Contrary to the previous table instruction-tuning work, we demonstrate that smaller learning rates and fewer training instances can enhance table understanding while preserving general capabilities. Based on our findings, we introduce TAMA, a TAble LLM instruction-tuned from LLaMA 3.1 8B Instruct, which achieves performance on par with, or surpassing GPT-3.5 and GPT-4 on table tasks, while maintaining strong out-of-domain generalization and general capabilities. Our findings highlight the potential for reduced data annotation costs and more efficient model development through careful hyperparameter selection. We open-source the project and our models.
Authors: Elham Kiyani, Khemraj Shukla, Jorge F. Urb\'an, J\'er\^ome Darbon, George Em Karniadakis
Abstract: Physics-Informed Neural Networks (PINNs) have revolutionized the computation of PDE solutions by integrating partial differential equations (PDEs) into the neural network's training process as soft constraints, becoming an important component of the scientific machine learning (SciML) ecosystem. More recently, physics-informed Kolmogorv-Arnold networks (PIKANs) have also shown to be effective and comparable in accuracy with PINNs. In their current implementation, both PINNs and PIKANs are mainly optimized using first-order methods like Adam, as well as quasi-Newton methods such as BFGS and its low-memory variant, L-BFGS. However, these optimizers often struggle with highly non-linear and non-convex loss landscapes, leading to challenges such as slow convergence, local minima entrapment, and (non)degenerate saddle points. In this study, we investigate the performance of Self-Scaled BFGS (SSBFGS), Self-Scaled Broyden (SSBroyden) methods and other advanced quasi-Newton schemes, including BFGS and L-BFGS with different line search strategies approaches. These methods dynamically rescale updates based on historical gradient information, thus enhancing training efficiency and accuracy. We systematically compare these optimizers -- using both PINNs and PIKANs -- on key challenging linear, stiff, multi-scale and non-linear PDEs, including the Burgers, Allen-Cahn, Kuramoto-Sivashinsky, and Ginzburg-Landau equations. Our findings provide state-of-the-art results with orders-of-magnitude accuracy improvements without the use of adaptive weights or any other enhancements typically employed in PINNs. More broadly, our results reveal insights into the effectiveness of second-order optimization strategies in significantly improving the convergence and accurate generalization of PINNs and PIKANs.
Authors: Han Fang, Paul Weng, Yutong Ban
Abstract: Recently, deep reinforcement learning (DRL) has achieved promising results in solving online 3D Bin Packing Problems (3D-BPP). However, these DRL-based policies may perform poorly on new instances due to distribution shift. Besides generalization, we also consider adaptation, completely overlooked by previous work, which aims at rapidly fine-tuning these policies to a new test distribution. To tackle both generalization and adaptation issues, we propose ASAP, which decomposes a solver's decision-making into two policies, one for proposal and one for selection. The role of the proposal policy is to suggest promising actions, which allows the selection policy to choose among them. To effectively learn these policies, we introduce a training framework that combines pre-training and post-training, enhanced by meta-learning. During online adaptation, we only fine-tune the selection policy to rapidly adapt to a test distribution. Our experiments demonstrate that ASAP exhibits excellent generalization and adaptation capabilities on in-distribution and out-of-distribution instances for both discrete and continuous setups.
Authors: Sangyeon Park, Isaac Han, Seungwon Oh, Kyung-Joong Kim
Abstract: Plasticity loss, a critical challenge in neural network training, limits a model's ability to adapt to new tasks or shifts in data distribution. This paper introduces AID (Activation by Interval-wise Dropout), a novel method inspired by Dropout, designed to address plasticity loss. Unlike Dropout, AID generates subnetworks by applying Dropout with different probabilities on each preactivation interval. Theoretical analysis reveals that AID regularizes the network, promoting behavior analogous to that of deep linear networks, which do not suffer from plasticity loss. We validate the effectiveness of AID in maintaining plasticity across various benchmarks, including continual learning tasks on standard image classification datasets such as CIFAR10, CIFAR100, and TinyImageNet. Furthermore, we show that AID enhances reinforcement learning performance in the Arcade Learning Environment benchmark.
Authors: Federico Malato, Ville Hautamaki
Abstract: Sample inefficiency is a long-lasting challenge in deep reinforcement learning (DRL). Despite dramatic improvements have been made, the problem is far from being solved and is especially challenging in environments with sparse or delayed rewards. In our work, we propose to use Adversarial Estimates as a new, simple and efficient approach to mitigate this problem for a class of feedback-based DRL algorithms. Our approach leverages latent similarity search from a small set of human-collected trajectories to boost learning, using only five minutes of human-recorded experience. The results of our study show algorithms trained with Adversarial Estimates converge faster than their original version. Moreover, we discuss how our approach could enable learning in feedback-based algorithms in extreme scenarios with very sparse rewards.
Authors: Oscar Skean, Md Rifat Arefin, Dan Zhao, Niket Patel, Jalal Naghiyev, Yann LeCun, Ravid Shwartz-Ziv
Abstract: From extracting features to generating text, the outputs of large language models (LLMs) typically rely on the final layers, following the conventional wisdom that earlier layers capture only low-level cues. However, our analysis shows that intermediate layers can encode even richer representations, often improving performance on a range of downstream tasks. To explain and quantify these hidden-layer properties, we propose a unified framework of representation quality metrics based on information theory, geometry, and invariance to input perturbations. Our framework highlights how each layer balances information compression and signal preservation, revealing why mid-depth embeddings can exceed the last layer's performance. Through extensive experiments on 32 text-embedding tasks across various architectures (transformers, state-space models) and domains (language, vision), we demonstrate that intermediate layers consistently provide stronger features, challenging the standard view on final-layer embeddings and opening new directions on using mid-layer representations for more robust and accurate representations.
Authors: Amin Heyrani Nobari, Kaveh Alimohammadi, Ali ArjomandBigdeli, Akash Srivastava, Faez Ahmed, Navid Azizan
Abstract: Model merging, a method that combines the parameters and embeddings of multiple fine-tuned large language models (LLMs), offers a promising approach to enhance model performance across various tasks while maintaining computational efficiency. This paper introduces Activation-Informed Merging (AIM), a technique that integrates the information from the activation space of LLMs into the merging process to improve performance and robustness. AIM is designed as a flexible, complementary solution that is applicable to any existing merging method. It aims to preserve critical weights from the base model, drawing on principles from continual learning (CL) and model compression. Utilizing a task-agnostic calibration set, AIM selectively prioritizes essential weights during merging. We empirically demonstrate that AIM significantly enhances the performance of merged models across multiple benchmarks. Our findings suggest that considering the activation-space information can provide substantial advancements in the model merging strategies for LLMs, with up to a 40% increase in benchmark performance.
Authors: Maohao Shen, Guangtao Zeng, Zhenting Qi, Zhang-Wei Hong, Zhenfang Chen, Wei Lu, Gregory Wornell, Subhro Das, David Cox, Chuang Gan
Abstract: Large language models (LLMs) have demonstrated remarkable reasoning capabilities across diverse domains. Recent studies have shown that increasing test-time computation enhances LLMs' reasoning capabilities. This typically involves extensive sampling at inference time guided by an external LLM verifier, resulting in a two-player system. Despite external guidance, the effectiveness of this system demonstrates the potential of a single LLM to tackle complex tasks. Thus, we pose a new research problem: Can we internalize the searching capabilities to fundamentally enhance the reasoning abilities of a single LLM? This work explores an orthogonal direction focusing on post-training LLMs for autoregressive searching (i.e., an extended reasoning process with self-reflection and self-exploration of new strategies). To achieve this, we propose the Chain-of-Action-Thought (COAT) reasoning and a two-stage training paradigm: 1) a small-scale format tuning stage to internalize the COAT reasoning format and 2) a large-scale self-improvement stage leveraging reinforcement learning. Our approach results in Satori, a 7B LLM trained on open-source models and data. Extensive empirical evaluations demonstrate that Satori achieves state-of-the-art performance on mathematical reasoning benchmarks while exhibits strong generalization to out-of-domain tasks. Code, data, and models are fully open-sourced.
Authors: Guogang Zhu, Xuefeng Liu, Jianwei Niu, Shaojie Tang, Xinghao Wu
Abstract: It is often observed that the aggregated model in FL underperforms on local data until after several rounds of local training. This temporary performance drop can potentially slow down the convergence of the FL model. Prior work regards this performance drop as an inherent cost of knowledge sharing among clients and does not give it special attention. While some studies directly focus on designing techniques to alleviate the issue, its root causes remain poorly understood. To bridge this gap, we construct a framework that enables layer-peeled analysis of how feature representations evolve during model aggregation in FL. It focuses on two key aspects: (1) the intrinsic quality of extracted features, and (2) the alignment between features and their subsequent parameters -- both of which are critical to downstream performance. Using this framework, we first investigate how model aggregation affects internal feature extraction process. Our analysis reveals that aggregation degrades feature quality and weakens the coupling between intermediate features and subsequent layers, both of which are well shaped during local training. More importantly, this degradation is not confined to specific layers but progressively accumulates with network depth -- a phenomenon we term Cumulative Feature Degradation (CFD). CFD significantly impairs the quality of penultimate-layer features and weakens their coupling with the classifier, ultimately degrading model performance. We further revisit several widely adopted solutions through the lens of layer-peeled feature extraction to understand why they are effective in addressing aggregation-induced performance drop. Our results show that their effectiveness lies in mitigating the feature degradation described above, which is well aligned with our observations.
Authors: Rui Pan, Boyao Wang, Shizhe Diao, Xingyuan Pan, Jipeng Zhang, Renjie Pi, Tong Zhang
Abstract: Small language models (SLMs) have attracted considerable attention from both academia and industry due to their broad range of applications in edge devices. To obtain SLMs with strong performance, conventional approaches either pre-train the models from scratch, which incurs substantial computational costs, or compress/prune existing large language models (LLMs), which results in performance drops and falls short in comparison to pre-training. In this paper, we investigate the family of acceleration methods that involve both structured pruning and model training. We found 1) layer-wise adaptive pruning (Adapt-Pruner) is extremely effective in LLMs and yields significant improvements over existing pruning techniques, 2) adaptive pruning equipped with further training leads to models comparable to those pre-training from scratch, 3) incremental pruning brings non-trivial performance gain by interleaving pruning with training and only removing a small portion of neurons ($\sim$5%) at a time. Experimental results on LLaMA-3.1-8B demonstrate that Adapt-Pruner outperforms conventional pruning methods, such as LLM-Pruner, FLAP, and SliceGPT, by an average of 1%-7% in accuracy on commonsense benchmarks. Additionally, Adapt-Pruner restores the performance of MobileLLM-125M to 600M on the MMLU benchmark with 200$\times$ fewer tokens via pruning from its larger counterparts, and discovers a new 1B model that surpasses LLaMA-3.2-1B in multiple benchmarks. The official code is released at https://github.com/research4pan/AdaptPruner.
Authors: Yik Siu Chan, Narutatsu Ri, Yuxin Xiao, Marzyeh Ghassemi
Abstract: Despite extensive safety alignment efforts, large language models (LLMs) remain vulnerable to jailbreak attacks that elicit harmful behavior. While existing studies predominantly focus on attack methods that require technical expertise, two critical questions remain underexplored: (1) Are jailbroken responses truly useful in enabling average users to carry out harmful actions? (2) Do safety vulnerabilities exist in more common, simple human-LLM interactions? In this paper, we demonstrate that LLM responses most effectively facilitate harmful actions when they are both actionable and informative--two attributes easily elicited in multi-step, multilingual interactions. Using this insight, we propose HarmScore, a jailbreak metric that measures how effectively an LLM response enables harmful actions, and Speak Easy, a simple multi-step, multilingual attack framework. Notably, by incorporating Speak Easy into direct request and jailbreak baselines, we see an average absolute increase of 0.319 in Attack Success Rate and 0.426 in HarmScore in both open-source and proprietary LLMs across four safety benchmarks. Our work reveals a critical yet often overlooked vulnerability: Malicious users can easily exploit common interaction patterns for harmful intentions.
Authors: Weihua Du, Yiming Yang, Sean Welleck
Abstract: Multi-sample aggregation strategies, such as majority voting and best-of-N sampling, are widely used in contemporary large language models (LLMs) to enhance predictive accuracy across various tasks. A key challenge in this process is temperature selection, which significantly impacts model performance. Existing approaches either rely on a fixed default temperature or require labeled validation data for tuning, which are often scarce and difficult to obtain. This paper addresses the challenge of automatically identifying the (near)-optimal temperature for different LLMs using multi-sample aggregation strategies, without relying on task-specific validation data. We provide a comprehensive analysis of temperature's role in performance optimization, considering variations in model architectures, datasets, task types, model sizes, and predictive accuracy. Furthermore, we propose a novel entropy-based metric for automated temperature optimization, which consistently outperforms fixed-temperature baselines. Additionally, we incorporate a stochastic process model to enhance interpretability, offering deeper insights into the relationship between temperature and model performance.
Authors: Max Geier, Khachatur Nazaryan, Timothy Zaklama, Liang Fu
Abstract: The attention mechanism has transformed artificial intelligence research by its ability to learn relations between objects. In this work, we explore how a many-body wavefunction ansatz constructed from a large-parameter self-attention neural network can be used to solve the interacting electron problem in solids. By a systematic neural-network variational Monte Carlo study on a moir\'e quantum material, we demonstrate that the self-attention ansatz provides an accurate and efficient solution without human bias. Moreover, our numerical study finds that the required number of variational parameters scales roughly as $N^2$ with the number of electrons, which opens a path towards efficient large-scale simulations.
Authors: Zicheng Liu, Siyuan Li, Zhiyuan Chen, Fang Wu, Chang Yu, Qirong Yang, Yucheng Guo, Yujie Yang, Xiaoming Zhang, Stan Z. Li
Abstract: The interactions between DNA, RNA, and proteins are fundamental to biological processes, as illustrated by the central dogma of molecular biology. Although modern biological pre-trained models have achieved great success in analyzing these macromolecules individually, their interconnected nature remains underexplored. This paper follows the guidance of the central dogma to redesign both the data and model pipeline and offers a comprehensive framework, Life-Code, that spans different biological functions. As for data flow, we propose a unified pipeline to integrate multi-omics data by reverse-transcribing RNA and reverse-translating amino acids into nucleotide-based sequences. As for the model, we design a codon tokenizer and a hybrid long-sequence architecture to encode the interactions between coding and non-coding regions through masked modeling pre-training. To model the translation and folding process with coding sequences, Life-Code learns protein structures of the corresponding amino acids by knowledge distillation from off-the-shelf protein language models. Such designs enable Life-Code to capture complex interactions within genetic sequences, providing a more comprehensive understanding of multi-omics with the central dogma. Extensive experiments show that Life-Code achieves state-of-the-art results on various tasks across three omics, highlighting its potential for advancing multi-omics analysis and interpretation.
Authors: Ai Chen, Yuxu Lu, Dong Yang, Junlin Zhou, Yan Fu, Duanbing Chen
Abstract: Salient object detection (SOD) plays a critical role in Intelligent Imaging, facilitating the detection and segmentation of key visual elements in an image. However, adverse imaging conditions such as haze during the day, low light, and haze at night severely degrade image quality and hinder reliable object detection in real-world scenarios. To address these challenges, we propose a multi-knowledge-oriented nighttime haze imaging enhancer (MKoIE), which integrates three tasks: daytime dehazing, low-light enhancement, and nighttime dehazing. The MKoIE incorporates two key innovative components: First, the network employs a task-oriented node learning mechanism to handle three specific degradation types: day-time haze, low light, and night-time haze conditions, with an embedded self-attention module enhancing its performance in nighttime imaging. In addition, multi-receptive field enhancement module that efficiently extracts multi-scale features through three parallel depthwise separable convolution branches with different dilation rates, capturing comprehensive spatial information with minimal computational overhead to meet the requirements of real-time imaging deployment. To ensure optimal image reconstruction quality and visual characteristics, we suggest a hybrid loss function. Extensive experiments on different types of weather/imaging conditions illustrate that MKoIE surpasses existing methods, enhancing the reliability, accuracy, and operational efficiency of intelligent imaging.
Authors: Ziyan Wang, Sizhe Wei, Xiaoming Huo, Hao Wang
Abstract: Diffusion models have made significant advancements in recent years. However, their performance often deteriorates when trained or fine-tuned on imbalanced datasets. This degradation is largely due to the disproportionate representation of majority and minority data in image-text pairs. In this paper, we propose a general fine-tuning approach, dubbed PoGDiff, to address this challenge. Rather than directly minimizing the KL divergence between the predicted and ground-truth distributions, PoGDiff replaces the ground-truth distribution with a Product of Gaussians (PoG), which is constructed by combining the original ground-truth targets with the predicted distribution conditioned on a neighboring text embedding. Experiments on real-world datasets demonstrate that our method effectively addresses the imbalance problem in diffusion models, improving both generation accuracy and quality.
Authors: Blanca Calvo Figueras, Eneko Sagarzazu, Julen Etxaniz, Jeremy Barnes, Pablo Gamallo, Iria De Dios Flores, Rodrigo Agerri
Abstract: We introduce a professionally translated extension of the TruthfulQA benchmark designed to evaluate truthfulness in Basque, Catalan, Galician, and Spanish. Truthfulness evaluations of large language models (LLMs) have primarily been conducted in English. However, the ability of LLMs to maintain truthfulness across languages remains under-explored. Our study evaluates 12 state-of-the-art open LLMs, comparing base and instruction-tuned models using human evaluation, multiple-choice metrics, and LLM-as-a-Judge scoring. Our findings reveal that, while LLMs perform best in English and worst in Basque (the lowest-resourced language), overall truthfulness discrepancies across languages are smaller than anticipated. Furthermore, we show that LLM-as-a-Judge correlates more closely with human judgments than multiple-choice metrics, and that informativeness plays a critical role in truthfulness assessment. Our results also indicate that machine translation provides a viable approach for extending truthfulness benchmarks to additional languages, offering a scalable alternative to professional translation. Finally, we observe that universal knowledge questions are better handled across languages than context- and time-dependent ones, highlighting the need for truthfulness evaluations that account for cultural and temporal variability. Dataset and code are publicly available under open licenses.
Authors: Yung-Sung Chuang, Benjamin Cohen-Wang, Shannon Zejiang Shen, Zhaofeng Wu, Hu Xu, Xi Victoria Lin, James Glass, Shang-Wen Li, Wen-tau Yih
Abstract: We introduce SelfCite, a novel self-supervised approach that aligns LLMs to generate high-quality, fine-grained, sentence-level citations for the statements in their generated responses. Instead of only relying on costly and labor-intensive annotations, SelfCite leverages a reward signal provided by the LLM itself through context ablation: If a citation is necessary, removing the cited text from the context should prevent the same response; if sufficient, retaining the cited text alone should preserve the same response. This reward can guide the inference-time best-of-N sampling strategy to improve citation quality significantly, as well as be used in preference optimization to directly fine-tune the models for generating better citations. The effectiveness of SelfCite is demonstrated by increasing citation F1 up to 5.3 points on the LongBench-Cite benchmark across five long-form question answering tasks. The source code is available at https://github.com/facebookresearch/SelfCite
Authors: Kenan Jiang, Li Xiong, Fei Liu
Abstract: We investigate factors contributing to LLM agents' success in competitive multi-agent environments, using auctions as a testbed where agents bid to maximize profit. The agents are equipped with bidding domain knowledge, distinct personas that reflect item preferences, and a memory of auction history. Our work extends the classic auction scenario by creating a realistic environment where multiple agents bid on houses, weighing aspects such as size, location, and budget to secure the most desirable homes at the lowest prices. Particularly, we investigate three key questions: (a) How does a persona influence an agent's behavior in a competitive setting? (b) Can an agent effectively profile its competitors' behavior during auctions? (c) How can persona profiling be leveraged to create an advantage using strategies such as theory of mind? Through a series of experiments, we analyze the behaviors of LLM agents and shed light on new findings. Our testbed, called HARBOR, offers a valuable platform for deepening our understanding of multi-agent workflows in competitive environments.
Authors: Haoyu Lei, Kaiwen Zhou, Yinchuan Li, Zhitang Chen, Farzan Farnia
Abstract: Diffusion-based Neural Combinatorial Optimization (NCO) has demonstrated effectiveness in solving NP-complete (NPC) problems by learning discrete diffusion models for solution generation, eliminating hand-crafted domain knowledge. Despite their success, existing NCO methods face significant challenges in both cross-scale and cross-problem generalization, and high training costs compared to traditional solvers. While recent studies on diffusion models have introduced training-free guidance approaches that leverage pre-defined guidance functions for conditional generation, such methodologies have not been extensively explored in combinatorial optimization. To bridge this gap, we propose a training-free inference time adaptation framework (DIFU-Ada) that enables both the zero-shot cross-problem transfer and cross-scale generalization capabilities of diffusion-based NCO solvers without requiring additional training. We provide theoretical analysis that helps understanding the cross-problem transfer capability. Our experimental results demonstrate that a diffusion solver, trained exclusively on the Traveling Salesman Problem (TSP), can achieve competitive zero-shot transfer performance across different problem scales on TSP variants, such as Prize Collecting TSP (PCTSP) and the Orienteering Problem (OP), through inference time adaptation.
Authors: Shuai Niu, Jing Ma, Hongzhan Lin, Liang Bai, Zhihua Wang, Wei Bi, Yida Xu, Guo Li, Xian Yang
Abstract: Large language models (LLMs) have shown remarkable performance in vision-language tasks, but their application in the medical field remains underexplored, particularly for integrating structured time series data with unstructured clinical notes. In clinical practice, dynamic time series data, such as lab test results, capture critical temporal patterns, while clinical notes provide rich semantic context. Merging these modalities is challenging due to the inherent differences between continuous signals and discrete text. To bridge this gap, we introduce ProMedTS, a novel self-supervised multimodal framework that employs prompt-guided learning to unify these heterogeneous data types. Our approach leverages lightweight anomaly detection to generate anomaly captions that serve as prompts, guiding the encoding of raw time series data into informative prompt embeddings. These prompt embeddings are aligned with textual representations in a shared latent space, preserving fine-grained temporal nuances alongside semantic insights. Furthermore, our framework incorporates tailored self-supervised objectives to enhance both intra- and inter-modal alignment. We evaluate ProMedTS on disease diagnosis tasks using real-world datasets, and the results demonstrate that our method consistently outperforms state-of-the-art approaches.
Authors: Peter Carragher, Abhinand Jha, R Raghav, Kathleen M. Carley
Abstract: Large Language Models (LLMs) demonstrate remarkable capabilities in question answering (QA), but metrics for assessing their reliance on memorization versus retrieval remain underdeveloped. Moreover, while finetuned models are state-of-the-art on closed-domain tasks, general-purpose models like GPT-4o exhibit strong zero-shot performance. This raises questions about the trade-offs between memorization, generalization, and retrieval. In this work, we analyze the extent to which multimodal retrieval-augmented VLMs memorize training data compared to baseline VLMs. Using the WebQA benchmark, we contrast finetuned models with baseline VLMs on multihop retrieval and question answering, examining the impact of finetuning on data memorization. To quantify memorization in end-to-end retrieval and QA systems, we propose several proxy metrics by investigating instances where QA succeeds despite retrieval failing. In line with existing work, we find that finetuned models rely more heavily on memorization than retrieval-augmented VLMs, and achieve higher accuracy as a result (72% vs 52% on WebQA test set). Finally, we present the first empirical comparison of the parametric effect between text and visual modalities. Here, we find that image-based questions have parametric response rates that are consistently 15-25% higher than for text-based questions in the WebQA dataset. As such, our measures pose a challenge for future work, both to account for differences in model memorization across different modalities and more generally to reconcile memorization and generalization in joint Retrieval-QA tasks.
Authors: Xun Deng, Han Zhong, Rui Ai, Fuli Feng, Zheng Wang, Xiangnan He
Abstract: Direct Preference Optimization (DPO) has emerged as a promising approach for aligning large language models with human preferences. While prior work mainly extends DPO from the aspect of the objective function, we instead improve DPO from the largely overlooked but critical aspect of data selection. Specifically, we address the issue of parameter shrinkage caused by noisy data by proposing a novel margin-maximization principle for dataset curation in DPO training. To further mitigate the noise in different reward models, we propose a Bayesian Aggregation approach that unifies multiple margin sources (external and implicit) into a single preference probability. Extensive experiments in diverse settings demonstrate the consistently high data efficiency of our approach. Remarkably, by using just 10\% of the Ultrafeedback dataset, our approach achieves 3\% to 8\% improvements across various Llama, Mistral, and Qwen models on the AlpacaEval2 benchmark. Furthermore, our approach seamlessly extends to iterative DPO, yielding a roughly 3\% improvement with 25\% online data, revealing the high redundancy in this presumed high-quality data construction manner. These results highlight the potential of data selection strategies for advancing preference optimization.
Authors: Kavosh Asadi, Julien Han, Idan Pipano, Xingzi Xu, Dominique Perrault-Joncas, Shoham Sabach, Karim Bouyarmane, Mohammad Ghavamzadeh
Abstract: Direct preference optimization (\texttt{DPO}) has emerged as a promising approach for solving the alignment problem in AI. In this paper, we make two counter-intuitive observations about \texttt{DPO}. First, we show that \texttt{DPO} loss could be derived by starting from an alternative optimization problem that only defines the KL guardrail on in-sample responses, unlike the original RLHF problem where guardrails are defined on the entire distribution. Second, we prove a surprising property of this alternative optimization problem, namely that under its optimal policy, both preferred and rejected responses tend to decrease in probability, a phenomenon typically displayed by DPO in practice. To control this behavior, we propose a set of constraints designed to limit the displacement of probability mass between the preferred and rejected responses in the reference and target policies. The resulting algorithm, which we call Constrained Controlled DPO (\texttt{C2-DPO}), has a meaningful RLHF interpretation. By hedging against the displacement, \texttt{C2-DPO} provides practical improvements over vanilla \texttt{DPO} when aligning several language models using standard preference datasets.
Authors: Tomer Raz, Michael Shalyt, Elyasheev Leibtag, Rotem Kalisch, Shachar Weinbaum, Yaron Hadad, Ido Kaminer
Abstract: The constant $\pi$ has fascinated scholars throughout the centuries, inspiring numerous formulas for its evaluation, such as infinite sums and continued fractions. Despite their individual significance, many of the underlying connections among formulas remain unknown, missing unifying theories that could unveil deeper understanding. The absence of a unifying theory reflects a broader challenge across math and science: knowledge is typically accumulated through isolated discoveries, while deeper connections often remain hidden. In this work, we present an automated framework for the unification of mathematical formulas. Our system combines large language models (LLMs) for systematic formula harvesting, an LLM-code feedback loop for validation, and a novel symbolic algorithm for clustering and eventual unification. We demonstrate this methodology on the hallmark case of $\pi$, an ideal testing ground for symbolic unification. Applying this approach to 455,050 arXiv papers, we validate 407 distinct formulas for $\pi$ and prove relations between 381 (94%) of them, of which 188 (46%) can be derived from a single mathematical object$\unicode{x2014}$linking canonical formulas by Euler, Gauss, Brouncker, and newer ones from algorithmic discoveries by the Ramanujan Machine. Our method generalizes to other constants, including $e$, $\zeta(3)$, and Catalan's constant, demonstrating the potential of AI-assisted mathematics to uncover hidden structures and unify knowledge across domains.
Authors: Siqi Ouyang, Xi Xu, Lei Li
Abstract: Simultaneous translation of unbounded streaming speech remains a challenging problem due to the need for effectively processing the history speech context and past translations so that quality and latency, including computation overhead, can be balanced. Most prior works assume pre-segmented speech, limiting their real-world applicability. In this paper, we propose InfiniSST, a novel approach that formulates SST as a multi-turn dialogue task, enabling seamless translation of unbounded speech. We construct translation trajectories and robust segments from MuST-C with multi-latency augmentation during training and develop a key-value (KV) cache management strategy to facilitate efficient inference. Experiments on MuST-C En-Es, En-De, and En-Zh demonstrate that InfiniSST reduces computation-aware latency by 0.5 to 1 second while maintaining the same translation quality compared to baselines. Ablation studies further validate the contributions of our data construction and cache management strategy. We release the code and demo at https://github.com/LeiLiLab/InfiniSST
Authors: Xiaobo Xia, Xiaofeng Liu, Jiale Liu, Kuai Fang, Lu Lu, Samet Oymak, William S. Currie, Tongliang Liu
Abstract: Water quality is foundational to environmental sustainability, ecosystem resilience, and public health. Deep learning models, particularly Long Short-Term Memory (LSTM) networks, offer transformative potential for large-scale water quality prediction and scientific insights generation. However, their widespread adoption in high-stakes decision-making, such as pollution mitigation and equitable resource allocation, is prevented by unresolved trustworthiness challenges including fairness, uncertainty, interpretability, robustness, generalizability, and reproducibility. In this work, we present the first comprehensive evaluation of trustworthiness in a continental-scale multi-task LSTM model predicting 20 water quality variables (encompassing physical/chemical processes, geochemical weathering, and nutrient cycling) across 482 U.S. basins. Our investigation uncovers systematic patterns of model performance disparities linked to basin characteristics, the inherent complexity of biogeochemical processes, and variable predictability, emphasizing critical performance fairness concerns. We further propose methodological frameworks for quantitatively evaluating critical aspects of trustworthiness, including uncertainty, interpretability, and robustness, identifying key limitations that could challenge reliable real-world deployment. This work serves as a timely call to action for advancing trustworthy data-driven methods for water resources management and provides a pathway to offering critical insights for researchers, decision-makers, and practitioners seeking to leverage artificial intelligence (AI) responsibly in environmental management.
Authors: Nicholas Roberts, Niladri Chatterji, Sharan Narang, Mike Lewis, Dieuwke Hupkes
Abstract: Scaling laws are a critical component of the LLM development pipeline, most famously as a way to forecast training decisions such as 'compute-optimally' trading-off parameter count and dataset size, alongside a more recent growing list of other crucial decisions. In this work, we ask whether compute-optimal scaling behaviour can be skill-dependent. In particular, we examine knowledge and reasoning-based skills such as knowledge-based QA and code generation, and we answer this question in the affirmative: scaling laws are skill-dependent. Next, to understand whether skill-dependent scaling is an artefact of the pretraining datamix, we conduct an extensive ablation of different datamixes and find that, also when correcting for datamix differences, knowledge and code exhibit fundamental differences in scaling behaviour. We conclude with an analysis of how our findings relate to standard compute-optimal scaling using a validation set, and find that a misspecified validation set can impact compute-optimal parameter count by nearly 50%, depending on its skill composition.
Authors: Jiachen Zhu, Xinlei Chen, Kaiming He, Yann LeCun, Zhuang Liu
Abstract: Normalization layers are ubiquitous in modern neural networks and have long been considered essential. This work demonstrates that Transformers without normalization can achieve the same or better performance using a remarkably simple technique. We introduce Dynamic Tanh (DyT), an element-wise operation $DyT($x$) = \tanh(\alpha $x$)$, as a drop-in replacement for normalization layers in Transformers. DyT is inspired by the observation that layer normalization in Transformers often produces tanh-like, $S$-shaped input-output mappings. By incorporating DyT, Transformers without normalization can match or exceed the performance of their normalized counterparts, mostly without hyperparameter tuning. We validate the effectiveness of Transformers with DyT across diverse settings, ranging from recognition to generation, supervised to self-supervised learning, and computer vision to language models. These findings challenge the conventional understanding that normalization layers are indispensable in modern neural networks, and offer new insights into their role in deep networks.
Authors: Qizhi Pei, Lijun Wu, Zhuoshi Pan, Yu Li, Honglin Lin, Chenlin Ming, Xin Gao, Conghui He, Rui Yan
Abstract: Large Language Models (LLMs) have shown impressive progress in mathematical reasoning. While data augmentation is promising to enhance mathematical problem-solving ability, current approaches are predominantly limited to instance-level modifications-such as rephrasing or generating syntactic variations-which fail to capture and leverage the intrinsic relational structures inherent in mathematical knowledge. Inspired by human learning processes, where mathematical proficiency develops through systematic exposure to interconnected concepts, we introduce MathFusion, a novel framework that enhances mathematical reasoning through cross-problem instruction synthesis. MathFusion implements this through three fusion strategies: (1) sequential fusion, which chains related problems to model solution dependencies; (2) parallel fusion, which combines analogous problems to reinforce conceptual understanding; and (3) conditional fusion, which creates context-aware selective problems to enhance reasoning flexibility. By applying these strategies, we generate a new dataset, \textbf{MathFusionQA}, followed by fine-tuning models (DeepSeekMath-7B, Mistral-7B, Llama3-8B) on it. Experimental results demonstrate that MathFusion achieves substantial improvements in mathematical reasoning while maintaining high data efficiency, boosting performance by 18.0 points in accuracy across diverse benchmarks while requiring only 45K additional synthetic instructions, representing a substantial improvement over traditional single-instruction approaches. Our datasets, models, and code are publicly available at https://github.com/QizhiPei/mathfusion.
Authors: Pengzhou Cheng, Zheng Wu, Zongru Wu, Aston Zhang, Zhuosheng Zhang, Gongshen Liu
Abstract: Autonomous graphical user interface (GUI) agents powered by multimodal large language models have shown great promise. However, a critical yet underexplored issue persists: over-execution, where the agent executes tasks in a fully autonomous way, without adequate assessment of its action confidence to compromise an adaptive human-agent collaboration. This poses substantial risks in complex scenarios, such as those involving ambiguous user instructions, unexpected interruptions, and environmental hijacks. To address the issue, we introduce OS-Kairos, an adaptive GUI agent capable of predicting confidence levels at each interaction step and efficiently deciding whether to act autonomously or seek human intervention. OS-Kairos is developed through two key mechanisms: (i) collaborative probing that annotates confidence scores at each interaction step; (ii) confidence-driven interaction that leverages these confidence scores to elicit the ability of adaptive interaction. Experimental results show that OS-Kairos substantially outperforms existing models on our curated dataset featuring complex scenarios, as well as on established benchmarks such as AITZ and Meta-GUI, with 24.59\%$\sim$87.29\% improvements in task success rate. OS-Kairos facilitates an adaptive human-agent collaboration, prioritizing effectiveness, generality, scalability, and efficiency for real-world GUI interaction. The dataset and codes are available at https://github.com/Wuzheng02/OS-Kairos.
Authors: Aneesh Vathul, Daniel Lee, Sheryl Chen, Arthi Tasmia
Abstract: Large Language Models (LLMs) have demonstrated impressive capabilities on a broad array of NLP tasks, but their tendency to produce hallucinations$\unicode{x2013}$plausible-sounding but factually incorrect content$\unicode{x2013}$poses severe challenges in high-stakes domains. Existing hallucination detection methods either bear the computational cost of multiple inference passes or sacrifice accuracy for efficiency with single-pass approaches, neither of which is ideal in resource-constrained environments such as edge devices. We propose the Shannon Entropy Distribution Hallucination Detector (ShED-HD), a novel hallucination detection framework that bridges this gap by classifying sequence-level entropy patterns using a lightweight BiLSTM architecture with single-headed attention. In contrast to prior approaches, ShED-HD efficiently detects distinctive uncertainty patterns across entire output sequences, preserving contextual awareness. Through in-depth evaluation on three datasets (BioASQ, TriviaQA, and Jeopardy Questions), we show that ShED-HD significantly outperforms other computationally efficient approaches in the out-of-distribution setting, while achieving comparable performance in the in-distribution setting. ShED-HD facilitates hallucination detection that is low-cost, accurate, and generalizable, improving the credibility of content generated by LLMs in resource-constrained environments where trustworthy AI functionality is crucial.
Authors: Siyin Wang, Wenyi Yu, Xianzhao Chen, Xiaohai Tian, Jun Zhang, Lu Lu, Yu Tsao, Junichi Yamagishi, Yuxuan Wang, Chao Zhang
Abstract: This paper explores a novel perspective to speech quality assessment by leveraging natural language descriptions, offering richer, more nuanced insights than traditional numerical scoring methods. Natural language feedback provides instructive recommendations and detailed evaluations, yet existing datasets lack the comprehensive annotations needed for this approach. To bridge this gap, we introduce QualiSpeech, a comprehensive low-level speech quality assessment dataset encompassing 11 key aspects and detailed natural language comments that include reasoning and contextual insights. Additionally, we propose the QualiSpeech Benchmark to evaluate the low-level speech understanding capabilities of auditory large language models (LLMs). Experimental results demonstrate that finetuned auditory LLMs can reliably generate detailed descriptions of noise and distortion, effectively identifying their types and temporal characteristics. The results further highlight the potential for incorporating reasoning to enhance the accuracy and reliability of quality assessments. The dataset will be released at https://huggingface.co/datasets/tsinghua-ee/QualiSpeech.
URLs: https://huggingface.co/datasets/tsinghua-ee/QualiSpeech.
Authors: Megan A. Brown, Shubham Atreja, Libby Hemphill, Patrick Y. Wu
Abstract: Researchers have proposed the use of generative large language models (LLMs) to label data for research and applied settings. This literature emphasizes the improved performance of these models relative to other natural language models, noting that generative LLMs typically outperform other models and even humans across several metrics. Previous literature has examined bias across many applications and contexts, but less work has focused specifically on bias in generative LLMs' responses to subjective annotation tasks. This bias could result in labels applied by LLMs that disproportionately align with majority groups over a more diverse set of viewpoints. In this paper, we evaluate how LLMs represent diverse viewpoints on these contentious tasks. Across four annotation tasks on four datasets, we show that LLMs do not show systematic substantial disagreement with annotators on the basis of demographics. Rather, we find that multiple LLMs tend to be biased in the same directions on the same demographic categories within the same datasets. Moreover, the disagreement between human annotators on the labeling task -- a measure of item difficulty -- is far more predictive of LLM agreement with human annotators. We conclude with a discussion of the implications for researchers and practitioners using LLMs for automated data annotation tasks. Specifically, we emphasize that fairness evaluations must be contextual, model choice alone will not solve potential issues of bias, and item difficulty must be integrated into bias assessments.
Authors: Gorjan Radevski, Kiril Gashteovski, Shahbaz Syed, Christopher Malon, Sebastien Nicolas, Chia-Chien Hung, Timo Sztyler, Verena Heu{\ss}er, Wiem Ben Rim, Masafumi Enomoto, Kunihiro Takeoka, Masafumi Oyamada, Goran Glava\v{s}, Carolin Lawrence
Abstract: Question Answering (QA) accounts for a significant portion of LLM usage "in the wild". However, LLMs sometimes produce false or misleading responses, also known as "hallucinations". Therefore, grounding the generated answers in contextually provided information -- i.e., providing evidence for the generated text -- is paramount for LLMs' trustworthiness. Providing this information is the task of context attribution. In this paper, we systematically study LLM-based approaches for this task, namely we investigate (i) zero-shot inference, (ii) LLM ensembling, and (iii) fine-tuning of small LMs on synthetic data generated by larger LLMs. Our key contribution is SynQA: a novel generative strategy for synthesizing context attribution data. Given selected context sentences, an LLM generates QA pairs that are supported by these sentences. This leverages LLMs' natural strengths in text generation while ensuring clear attribution paths in the synthetic training data. We show that the attribution data synthesized via SynQA is highly effective for fine-tuning small LMs for context attribution in different QA tasks and domains. Finally, with a user study, we validate the usefulness of small LMs (fine-tuned on synthetic data from SynQA) in context attribution for QA.
Authors: Markus Flicke, Glenn Angrabeit, Madhav Iyengar, Vitalii Protsenko, Illia Shakun, Jovan Cicvaric, Bora Kargi, Haoyu He, Lukas Schuler, Lewin Scholz, Kavyanjali Agnihotri, Yong Cao, Andreas Geiger
Abstract: Scholar Inbox is a new open-access platform designed to address the challenges researchers face in staying current with the rapidly expanding volume of scientific literature. We provide personalized recommendations, continuous updates from open-access archives (arXiv, bioRxiv, etc.), visual paper summaries, semantic search, and a range of tools to streamline research workflows and promote open research access. The platform's personalized recommendation system is trained on user ratings, ensuring that recommendations are tailored to individual researchers' interests. To further enhance the user experience, Scholar Inbox also offers a map of science that provides an overview of research across domains, enabling users to easily explore specific topics. We use this map to address the cold start problem common in recommender systems, as well as an active learning strategy that iteratively prompts users to rate a selection of papers, allowing the system to learn user preferences quickly. We evaluate the quality of our recommendation system on a novel dataset of 800k user ratings, which we make publicly available, as well as via an extensive user study. https://www.scholar-inbox.com/
Authors: Andreas Happe, J\"urgen Cito
Abstract: Large Language Models (LLMs) have emerged as a powerful approach for driving offensive penetration-testing tooling. Due to the opaque nature of LLMs, empirical methods are typically used to analyze their efficacy. The quality of this analysis is highly dependent on the chosen testbed, captured metrics and analysis methods employed. This paper analyzes the methodology and benchmarking practices used for evaluating Large Language Model (LLM)-driven attacks, focusing on offensive uses of LLMs in cybersecurity. We review 19 research papers detailing 18 prototypes and their respective testbeds. We detail our findings and provide actionable recommendations for future research, emphasizing the importance of extending existing testbeds, creating baselines, and including comprehensive metrics and qualitative analysis. We also note the distinction between security research and practice, suggesting that CTF-based challenges may not fully represent real-world penetration testing scenarios.
Authors: Minh-Anh Nguyen, Dung D. Le
Abstract: Language representation learning has emerged as a promising approach for sequential recommendation, thanks to its ability to learn generalizable representations. However, despite its advantages, this approach still struggles with data sparsity and a limited understanding of common-sense user preferences. To address these limitations, we propose $\textbf{JEPA4Rec}$, a framework that combines $\textbf{J}$oint $\textbf{E}$mbedding $\textbf{P}$redictive $\textbf{A}$rchitecture with language modeling of item textual descriptions. JEPA4Rec captures semantically rich and transferable representations, improving recommendation performance and reducing reliance on large-scale pre-training data. Specifically, JEPA4Rec represents items as text sentences by flattening descriptive information such as $\textit{title, category}$, and other attributes. To encode these sentences, we employ a bidirectional Transformer encoder with modified embedding layers tailored for capturing item information in recommendation datasets. We apply masking to text sentences and use them to predict the representations of the unmasked sentences, helping the model learn generalizable item embeddings. To further improve recommendation performance and language understanding, we employ a two-stage training strategy incorporating self-supervised learning losses. Experiments on six real-world datasets demonstrate that JEPA4Rec consistently outperforms state-of-the-art methods, particularly in cross-domain, cross-platform, and low-resource scenarios.
Authors: Dejan Stancevic, Florian Handke, Luca Ambrogioni
Abstract: The practical performance of generative diffusion models depends on the appropriate choice of the noise scheduling function, which can also be equivalently expressed as a time reparameterization. In this paper, we present a time scheduler that selects sampling points based on entropy rather than uniform time spacing, ensuring that each point contributes an equal amount of information to the final generation. We prove that this time reparameterization does not depend on the initial choice of time. Furthermore, we provide a tractable exact formula to estimate this \emph{entropic time} for a trained model using the training loss without substantial overhead. Alongside the entropic time, inspired by the optimality results, we introduce a rescaled entropic time. In our experiments with mixtures of Gaussian distributions and ImageNet, we show that using the (rescaled) entropic times greatly improves the inference performance of trained models. In particular, we found that the image quality in pretrained EDM2 models, as evaluated by FID and FD-DINO scores, can be substantially increased by the rescaled entropic time reparameterization without increasing the number of function evaluations, with greater improvements in the few NFEs regime.
Authors: David Almog
Abstract: There is growing enthusiasm about the potential for humans and AI to collaborate by leveraging their respective strengths. Yet in practice, this promise often falls short. This paper uses an online experiment to identify non-instrumental image concerns as a key reason individuals underutilize AI recommendations. I show that concerns about how one is perceived, even when those perceptions carry no monetary consequences, lead participants to disregard AI advice and reduce task performance.
Authors: Zhun Wang, Vincent Siu, Zhe Ye, Tianneng Shi, Yuzhou Nie, Xuandong Zhao, Chenguang Wang, Wenbo Guo, Dawn Song
Abstract: The strong planning and reasoning capabilities of Large Language Models (LLMs) have fostered the development of agent-based systems capable of leveraging external tools and interacting with increasingly complex environments. However, these powerful features also introduce a critical security risk: indirect prompt injection, a sophisticated attack vector that compromises the core of these agents, the LLM, by manipulating contextual information rather than direct user prompts. In this work, we propose a generic black-box fuzzing framework, AgentVigil, designed to automatically discover and exploit indirect prompt injection vulnerabilities across diverse LLM agents. Our approach starts by constructing a high-quality initial seed corpus, then employs a seed selection algorithm based on Monte Carlo Tree Search (MCTS) to iteratively refine inputs, thereby maximizing the likelihood of uncovering agent weaknesses. We evaluate AgentVigil on two public benchmarks, AgentDojo and VWA-adv, where it achieves 71% and 70% success rates against agents based on o3-mini and GPT-4o, respectively, nearly doubling the performance of baseline attacks. Moreover, AgentVigil exhibits strong transferability across unseen tasks and internal LLMs, as well as promising results against defenses. Beyond benchmark evaluations, we apply our attacks in real-world environments, successfully misleading agents to navigate to arbitrary URLs, including malicious sites.
Authors: Ziluo Ding, Haobin Jiang, Yuxuan Wang, Zhenguo Sun, Yu Zhang, Xiaojie Niu, Ming Yang, Weishuai Zeng, Xinrun Xu, Zongqing Lu
Abstract: This paper presents JAEGER, a dual-level whole-body controller for humanoid robots that addresses the challenges of training a more robust and versatile policy. Unlike traditional single-controller approaches, JAEGER separates the control of the upper and lower bodies into two independent controllers, so that they can better focus on their distinct tasks. This separation alleviates the dimensionality curse and improves fault tolerance. JAEGER supports both root velocity tracking (coarse-grained control) and local joint angle tracking (fine-grained control), enabling versatile and stable movements. To train the controller, we utilize a human motion dataset (AMASS), retargeting human poses to humanoid poses through an efficient retargeting network, and employ a curriculum learning approach. This method performs supervised learning for initialization, followed by reinforcement learning for further exploration. We conduct our experiments on two humanoid platforms and demonstrate the superiority of our approach against state-of-the-art methods in both simulation and real environments.
Authors: Prithwish Dan, Kushal Kedia, Angela Chao, Edward Weiyi Duan, Maximus Adrian Pace, Wei-Chiu Ma, Sanjiban Choudhury
Abstract: Human videos offer a scalable way to train robot manipulation policies, but lack the action labels needed by standard imitation learning algorithms. Existing cross-embodiment approaches try to map human motion to robot actions, but often fail when the embodiments differ significantly. We propose X-Sim, a real-to-sim-to-real framework that uses object motion as a dense and transferable signal for learning robot policies. X-Sim starts by reconstructing a photorealistic simulation from an RGBD human video and tracking object trajectories to define object-centric rewards. These rewards are used to train a reinforcement learning (RL) policy in simulation. The learned policy is then distilled into an image-conditioned diffusion policy using synthetic rollouts rendered with varied viewpoints and lighting. To transfer to the real world, X-Sim introduces an online domain adaptation technique that aligns real and simulated observations during deployment. Importantly, X-Sim does not require any robot teleoperation data. We evaluate it across 5 manipulation tasks in 2 environments and show that it: (1) improves task progress by 30% on average over hand-tracking and sim-to-real baselines, (2) matches behavior cloning with 10x less data collection time, and (3) generalizes to new camera viewpoints and test-time changes. Code and videos are available at https://portal-cornell.github.io/X-Sim/.
Authors: Emanuele La Malfa, Jon Vadillo, Marco Molinari, Michael Wooldridge
Abstract: This paper introduces a formal notion of fixed point explanations, inspired by the "why regress" principle, to assess, through recursive applications, the stability of the interplay between a model and its explainer. Fixed point explanations satisfy properties like minimality, stability, and faithfulness, revealing hidden model behaviours and explanatory weaknesses. We define convergence conditions for several classes of explainers, from feature-based to mechanistic tools like Sparse AutoEncoders, and we report quantitative and qualitative results.
Authors: Eric Hanchen Jiang, Haozheng Luo, Shengyuan Pang, Xiaomin Li, Zhenting Qi, Hengli Li, Cheng-Fu Yang, Zongyu Lin, Xinfeng Li, Hao Xu, Kai-Wei Chang, Ying Nian Wu
Abstract: Mathematical reasoning presents a significant challenge for Large Language Models (LLMs), often requiring robust multi step logical consistency. While Chain of Thought (CoT) prompting elicits reasoning steps, it doesn't guarantee correctness, and improving reliability via extensive sampling is computationally costly. This paper introduces the Energy Outcome Reward Model (EORM), an effective, lightweight, post hoc verifier. EORM leverages Energy Based Models (EBMs) to simplify the training of reward models by learning to assign a scalar energy score to CoT solutions using only outcome labels, thereby avoiding detailed annotations. It achieves this by interpreting discriminator output logits as negative energies, effectively ranking candidates where lower energy is assigned to solutions leading to correct final outcomes implicitly favoring coherent reasoning. On mathematical benchmarks (GSM8k, MATH), EORM significantly improves final answer accuracy (e.g., with Llama 3 8B, achieving 90.7% on GSM8k and 63.7% on MATH). EORM effectively leverages a given pool of candidate solutions to match or exceed the performance of brute force sampling, thereby enhancing LLM reasoning outcome reliability through its streamlined post hoc verification process.
Authors: Adrian Arnaiz-Rodriguez, Federico Errica
Abstract: After a renaissance phase in which researchers revisited the message-passing paradigm through the lens of deep learning, the graph machine learning community shifted its attention towards a deeper and practical understanding of message-passing's benefits and limitations. In this position paper, we notice how the fast pace of progress around the topics of oversmoothing and oversquashing, the homophily-heterophily dichotomy, and long-range tasks, came with the consolidation of commonly accepted beliefs and assumptions that are not always true nor easy to distinguish from each other. We argue that this has led to ambiguities around the investigated problems, preventing researchers from focusing on and addressing precise research questions while causing a good amount of misunderstandings. Our contribution wants to make such common beliefs explicit and encourage critical thinking around these topics, supported by simple but noteworthy counterexamples. The hope is to clarify the distinction between the different issues and promote separate but intertwined research directions to address them.
Authors: Chao Pang, Vincent Jeanselme, Young Sang Choi, Xinzhuo Jiang, Zilin Jing, Aparajita Kashyap, Yuta Kobayashi, Yanwei Li, Florent Pollet, Karthik Natarajan, Shalmali Joshi
Abstract: Foundation models hold significant promise in healthcare, given their capacity to extract meaningful representations independent of downstream tasks. This property has enabled state-of-the-art performance across several clinical applications trained on structured electronic health record (EHR) data, even in settings with limited labeled data, a prevalent challenge in healthcare. However, there is little consensus on these models' potential for clinical utility due to the lack of desiderata of comprehensive and meaningful tasks and sufficiently diverse evaluations to characterize the benefit over conventional supervised learning. To address this gap, we propose a suite of clinically meaningful tasks spanning patient outcomes, early prediction of acute and chronic conditions, including desiderata for robust evaluations. We evaluate state-of-the-art foundation models on EHR data consisting of 5 million patients from Columbia University Irving Medical Center (CUMC), a large urban academic medical center in New York City, across 14 clinically relevant tasks. We measure overall accuracy, calibration, and subpopulation performance to surface tradeoffs based on the choice of pre-training, tokenization, and data representation strategies. Our study aims to advance the empirical evaluation of structured EHR foundation models and guide the development of future healthcare foundation models.
Authors: Anna Spagnolli, Cecilia Tolomini, Elisa Beretta, Claudio Sarra
Abstract: Artificial Intelligence (AI) plays an essential role in healthcare and is pervasively incorporated into medical software and equipment. In the European Union, healthcare is a high-risk application domain for AI, and providers must prepare Instructions for Use (IFU) according to the European regulation 2024/1689 (AI Act). To this regulation, the principle of transparency is cardinal and requires the IFU to be clear and relevant to the users. This study tests whether these latter requirements are satisfied by the IFU structure. A survey was administered online via the Qualtrics platform to four types of direct stakeholders, i.e., managers (N = 238), healthcare professionals (N = 115), patients (N = 229), and Information Technology experts (N = 230). The participants rated the relevance of a set of transparency needs and indicated the IFU section addressing them. The results reveal differentiated priorities across stakeholders and a troubled mapping of transparency needs onto the IFU structure. Recommendations to build a locally meaningful IFU are derived.
Authors: Ansel Blume, Jeonghwan Kim, Hyeonjeong Ha, Elen Chatikyan, Xiaomeng Jin, Khanh Duy Nguyen, Nanyun Peng, Kai-Wei Chang, Derek Hoiem, Heng Ji
Abstract: Real-world objects are composed of distinctive, object-specific parts. Identifying these parts is key to performing fine-grained, compositional reasoning-yet, large multimodal models (LMMs) struggle to perform this seemingly straightforward task. In this work, we introduce PARTONOMY, an LMM benchmark designed for pixel-level part grounding. We construct PARTONOMY from existing part datasets and our own rigorously annotated set of images, encompassing 862 part labels and 534 object labels for evaluation. Unlike existing datasets that simply ask models to identify generic parts, PARTONOMY uses specialized concepts (e.g., agricultural airplane), and challenges models to compare objects' parts, consider part-whole relationships, and justify textual predictions with visual segmentations. Our experiments demonstrate significant limitations in state-of-the-art LMMs (e.g., LISA-13B achieves only 5.9% gIoU), highlighting a critical gap in their part grounding abilities. We note that existing segmentation-enabled LMMs (segmenting LMMs) have two key architectural shortcomings: they use special [SEG] tokens not seen during pretraining which induce distribution shift, and they discard predicted segmentations instead of using past predictions to guide future ones. To address these deficiencies, we train several part-centric LMMs and propose PLUM, a novel segmenting LMM that uses span tagging instead of segmentation tokens and that conditions on prior predictions in a feedback loop. We find that pretrained PLUM outperforms existing segmenting LMMs on reasoning segmentation, VQA, and visual hallucination benchmarks. In addition, PLUM finetuned on our proposed Explanatory Part Segmentation task is competitive with segmenting LMMs trained on significantly more segmentation data. Our work opens up new avenues towards enabling fine-grained, grounded visual understanding in LMMs.
Authors: Fengqing Jiang, Fengbo Ma, Zhangchen Xu, Yuetai Li, Bhaskar Ramasubramanian, Luyao Niu, Bo Li, Xianyan Chen, Zhen Xiang, Radha Poovendran
Abstract: Large language models (LLMs) exhibit advancing capabilities in complex tasks, such as reasoning and graduate-level question answering, yet their resilience against misuse, particularly involving scientifically sophisticated risks, remains underexplored. Existing safety benchmarks typically focus either on instructions requiring minimal knowledge comprehension (e.g., ``tell me how to build a bomb") or utilize prompts that are relatively low-risk (e.g., multiple-choice or classification tasks about hazardous content). Consequently, they fail to adequately assess model safety when handling knowledge-intensive, hazardous scenarios. To address this critical gap, we introduce SOSBench, a regulation-grounded, hazard-focused benchmark encompassing six high-risk scientific domains: chemistry, biology, medicine, pharmacology, physics, and psychology. The benchmark comprises 3,000 prompts derived from real-world regulations and laws, systematically expanded via an LLM-assisted evolutionary pipeline that introduces diverse, realistic misuse scenarios (e.g., detailed explosive synthesis instructions involving advanced chemical formulas). We evaluate frontier models within a unified evaluation framework using our SOSBench. Despite their alignment claims, advanced models consistently disclose policy-violating content across all domains, demonstrating alarmingly high rates of harmful responses (e.g., 79.1% for Deepseek-R1 and 47.3% for GPT-4.1). These results highlight significant safety alignment deficiencies and underscore urgent concerns regarding the responsible deployment of powerful LLMs.
Authors: Haomiao Qiu, Miao Zhang, Ziyue Qiao, Liqiang Nie
Abstract: Continual Learning (CL) aims to enable models to continuously acquire new knowledge from a sequence of tasks with avoiding the forgetting of learned information. However, existing CL methods only rely on the parameters of the most recent task for inference, which makes them susceptible to catastrophic forgetting. Inspired by the recent success of model merging techniques, we propose \textbf{Perturb-and-Merge (P\&M)}, a novel continual learning framework that integrates model merging into the CL paradigm to mitigate forgetting. Specifically, after training on each task, P\&M constructs a new model by forming a convex combination of the previous model and the newly trained task-specific model. Through theoretical analysis, we minimize the total loss increase across all tasks and derive an analytical solution for the optimal merging coefficient. To further improve the performance of the merged model, we observe that the degradation introduced during merging can be alleviated by a regularization term composed of the task vector and the Hessian matrix of the loss function. Interestingly, we show that this term can be efficiently approximated using second-order symmetric finite differences, and a stochastic perturbation strategy along the task vector direction is accordingly devised which incurs no additional forward or backward passes while providing an effective approximation of the regularization term. Finally, we combine P\&M with LoRA, a parameter-efficient fine-tuning method, to reduce memory overhead. Our proposed approach achieves state-of-the-art performance on several continual learning benchmark datasets.
Authors: Luca Maria Del Bono, Federico Ricci-Tersenghi, Francesco Zamponi
Abstract: Recent years have seen a rise in the application of machine learning techniques to aid the simulation of hard-to-sample systems that cannot be studied using traditional methods. Despite the introduction of many different architectures and procedures, a wide theoretical understanding is still lacking, with the risk of suboptimal implementations. As a first step to address this gap, we provide here a complete analytic study of the widely-used Sequential Tempering procedure applied to a shallow MADE architecture for the Curie-Weiss model. The contribution of this work is twofold: firstly, we give a description of the optimal weights and of the training under Gradient Descent optimization. Secondly, we compare what happens in Sequential Tempering with and without the addition of local Metropolis Monte Carlo steps. We are thus able to give theoretical predictions on the best procedure to apply in this case. This work establishes a clear theoretical basis for the integration of machine learning techniques into Monte Carlo sampling and optimization.
Authors: Dongwoo Lee, Dong Bok Lee, Steven Adriaensen, Juho Lee, Sung Ju Hwang, Frank Hutter, Seon Joo Kim, Hae Beom Lee
Abstract: Scaling has been a major driver of recent advancements in deep learning. Numerous empirical studies have found that scaling laws often follow the power-law and proposed several variants of power-law functions to predict the scaling behavior at larger scales. However, existing methods mostly rely on point estimation and do not quantify uncertainty, which is crucial for real-world applications involving decision-making problems such as determining the expected performance improvements achievable by investing additional computational resources. In this work, we explore a Bayesian framework based on Prior-data Fitted Networks (PFNs) for neural scaling law extrapolation. Specifically, we design a prior distribution that enables the sampling of infinitely many synthetic functions resembling real-world neural scaling laws, allowing our PFN to meta-learn the extrapolation. We validate the effectiveness of our approach on real-world neural scaling laws, comparing it against both the existing point estimation methods and Bayesian approaches. Our method demonstrates superior performance, particularly in data-limited scenarios such as Bayesian active learning, underscoring its potential for reliable, uncertainty-aware extrapolation in practical applications.
Authors: Giovanni Acampora, Andris Ambainis, Natalia Ares, Leonardo Banchi, Pallavi Bhardwaj, Daniele Binosi, G. Andrew D. Briggs, Tommaso Calarco, Vedran Dunjko, Jens Eisert, Olivier Ezratty, Paul Erker, Federico Fedele, Elies Gil-Fuster, Martin G\"arttner, Mats Granath, Markus Heyl, Iordanis Kerenidis, Matthias Klusch, Anton Frisk Kockum, Richard Kueng, Mario Krenn, J\"org L\"assig, Antonio Macaluso, Sabrina Maniscalco, Florian Marquardt, Kristel Michielsen, Gorka Mu\~noz-Gil, Daniel M\"ussig, Hendrik Poulsen Nautrup, Sophie A. Neubauer, Evert van Nieuwenburg, Roman Orus, J\"org Schmiedmayer, Markus Schmitt, Philipp Slusallek, Filippo Vicentini, Christof Weitenberg, Frank K. Wilhelm
Abstract: This white paper discusses and explores the various points of intersection between quantum computing and artificial intelligence (AI). It describes how quantum computing could support the development of innovative AI solutions. It also examines use cases of classical AI that can empower research and development in quantum technologies, with a focus on quantum computing and quantum sensing. The purpose of this white paper is to provide a long-term research agenda aimed at addressing foundational questions about how AI and quantum computing interact and benefit one another. It concludes with a set of recommendations and challenges, including how to orchestrate the proposed theoretical work, align quantum AI developments with quantum hardware roadmaps, estimate both classical and quantum resources - especially with the goal of mitigating and optimizing energy consumption - advance this emerging hybrid software engineering discipline, and enhance European industrial competitiveness while considering societal implications.
Authors: Zefan Cai, Wen Xiao, Hanshi Sun, Cheng Luo, Yikai Zhang, Ke Wan, Yucheng Li, Yeyang Zhou, Li-Wen Chang, Jiuxiang Gu, Zhen Dong, Anima Anandkumar, Abedelkadir Asi, Junjie Hu
Abstract: Reasoning models have demonstrated impressive performance in self-reflection and chain-of-thought reasoning. However, they often produce excessively long outputs, leading to prohibitively large key-value (KV) caches during inference. While chain-of-thought inference significantly improves performance on complex reasoning tasks, it can also lead to reasoning failures when deployed with existing KV cache compression approaches. To address this, we propose Redundancy-aware KV Cache Compression for Reasoning models (R-KV), a novel method specifically targeting redundant tokens in reasoning models. Our method preserves nearly 100% of the full KV cache performance using only 10% of the KV cache, substantially outperforming existing KV cache baselines, which reach only 60% of the performance. Remarkably, R-KV even achieves 105% of full KV cache performance with 16% of the KV cache. This KV-cache reduction also leads to a 90% memory saving and a 6.6X throughput over standard chain-of-thought reasoning inference. Experimental results show that R-KV consistently outperforms existing KV cache compression baselines across two mathematical reasoning datasets.
Authors: Chamika Sudusinghe, Gerasimos Gerogiannis, Damitha Lenadora, Charles Block, Josep Torrellas, Charith Mendis
Abstract: Sparse tensor programs are essential in deep learning and graph analytics, driving the need for optimized processing. To meet this demand, specialized hardware accelerators are being developed. Optimizing these programs for accelerators is challenging for two reasons: program performance is highly sensitive to variations in sparse inputs, and early-stage accelerators rely on expensive simulators. Therefore, ML-based cost models used for optimizing such programs on general-purpose hardware are often ineffective for early-stage accelerators, as they require large datasets for proper training. To this end, we introduce COGNATE, a novel framework that leverages inexpensive data samples from general-purpose hardware (e.g., CPUs) to train cost models, followed by few-shot fine-tuning on emerging hardware. COGNATE exploits the homogeneity of input features across hardware platforms while effectively mitigating heterogeneity, enabling cost model training with just 5% of the data samples needed by accelerator-specific models to achieve comparable performance. We conduct extensive experiments to demonstrate that COGNATE outperforms existing techniques, achieving average speedups of 1.47x (up to 5.46x) for SpMM and 1.39x (up to 4.22x) for SDDMM.
Authors: Kordel K. France, Ovidiu Daescu
Abstract: Robotic odour source localization (OSL) is a critical capability for autonomous systems operating in complex environments. However, current OSL methods often suffer from ambiguities, particularly when robots misattribute odours to incorrect objects due to limitations in olfactory datasets and sensor resolutions. To address this challenge, we introduce a novel machine learning method using diffusion-based molecular generation to enhance odour localization accuracy that can be used by itself or with automated olfactory dataset construction pipelines with vision-language models (VLMs) This generative process of our diffusion model expands the chemical space beyond the limitations of both current olfactory datasets and the training data of VLMs, enabling the identification of potential odourant molecules not previously documented. The generated molecules can then be more accurately validated using advanced olfactory sensors which emulate human olfactory recognition through electronic sensor arrays. By integrating visual analysis, language processing, and molecular generation, our framework enhances the ability of olfaction-vision models on robots to accurately associate odours with their correct sources, thereby improving navigation and decision-making through better sensor selection for a target compound. Our methodology represents a foundational advancement in the field of artificial olfaction, offering a scalable solution to the challenges posed by limited olfactory data and sensor ambiguities.
Authors: Andreu Ballus Santacana
Abstract: We present a tractable, incremental framework for topological dialogue semantics based on finite, discrete semantic spaces. Building on the intuition that utterances correspond to open sets and their combinatorial relations form a simplicial complex (the dialogue nerve), we give a rigorous foundation, a provably correct incremental algorithm for nerve updates, and a reference implementation in the Wolfram Language. The framework supports negative nerve computation (inconsistency tracking), consequence extraction, and a transparent, set-theoretic ranking of entailments. We clarify which combinatorial properties hold in the discrete case, provide motivating examples, and outline limitations and prospects for richer logical and categorical extensions.
Authors: Debarati Bhattacharjee, Ashish Anand
Abstract: This paper presents a framework to convert argumentative texts into argument knowledge graphs (AKG). Starting with basic annotations of argumentative components (ACs) and argumentative relations (ARs), we enrich the information by constructing a knowledge base (KB) graph with metadata attributes for nodes. Next, we use premises and inference rules from the KB to form arguments by applying modus ponens. From these arguments, we create an AKG. The nodes and edges of the AKG have attributes that capture important argumentative features. We also find missing inference rules by identifying markers. This makes it possible to identify undercut attacks that were previously undetectable in existing datasets. The AKG gives a graphical view of the argumentative structure that is easier to understand than theoretical formats. It also prepares the ground for future reasoning tasks, including checking the coherence of arguments and identifying opportunities for revision. For this, it is important to find indirect relations, many of which are implicit. Our proposed AKG format, with annotated inference rules and modus ponens, will help reasoning models learn the implicit indirect relations that require inference over arguments and the relations between them.
Authors: Yulei Qin, Gang Li, Zongyi Li, Zihan Xu, Yuchen Shi, Zhekai Lin, Xiao Cui, Ke Li, Xing Sun
Abstract: Existing large language models (LLMs) face challenges of following complex instructions, especially when multiple constraints are present and organized in paralleling, chaining, and branching structures. One intuitive solution, namely chain-of-thought (CoT), is expected to universally improve capabilities of LLMs. However, we find that the vanilla CoT exerts a negative impact on performance due to its superficial reasoning pattern of simply paraphrasing the instructions. It fails to peel back the compositions of constraints for identifying their relationship across hierarchies of types and dimensions. To this end, we propose a systematic method to boost LLMs in dealing with complex instructions via incentivizing reasoning for test-time compute scaling. First, we stem from the decomposition of complex instructions under existing taxonomies and propose a reproducible data acquisition method. Second, we exploit reinforcement learning (RL) with verifiable rule-centric reward signals to cultivate reasoning specifically for instruction following. We address the shallow, non-essential nature of reasoning under complex instructions via sample-wise contrast for superior CoT enforcement. We also exploit behavior cloning of experts to facilitate steady distribution shift from fast-thinking LLMs to skillful reasoners. Extensive evaluations on seven comprehensive benchmarks confirm the validity of the proposed method, where a 1.5B LLM achieves 11.74% gains with performance comparable to a 8B LLM. Codes and data will be available later (under review). Keywords: reinforcement learning with verifiable rewards (RLVR), instruction following, complex instructions
Authors: Jiahao Wang, Jinbo Han, Xingda Wei, Sijie Shen, Dingyan Zhang, Chenguang Fang, Rong Chen, Wenyuan Yu, Haibo Chen
Abstract: Serving large language models (LLMs) is important for cloud providers, and caching intermediate results (KV\$) after processing each request substantially improves serving throughput and latency. However, there is limited understanding of how LLM serving benefits from KV\$ caching, where system design decisions like cache eviction policies are highly workload-dependent. In this paper, we present the first systematic characterization of the KV\$ workload patterns from one of the leading LLM service providers. We draw observations that were not covered by previous studies focusing on synthetic workloads, including: KV\$ reuses are skewed across requests, where reuses between single-turn requests are equally important as multi-turn requests; the reuse time and probability are diverse considering all requests, but for a specific request category, the pattern tends to be predictable; and the overall cache size required for an ideal cache hit ratio is moderate. Based on the characterization, we further propose a workload-aware cache eviction policy that improves the serving performance under real-world traces, especially with limited cache capacity.
Authors: Isarun Chamveha, Supphanut Chaiyungyuen, Sasinun Worakriangkrai, Nattawadee Prasawang, Warasinee Chaisangmongkon, Pornpim Korpraphong, Voraparee Suvannarerg, Shanigarn Thiravit, Chalermdej Kannawat, Kewalin Rungsinaporn, Suwara Issaragrisil, Payia Chadbunchachai, Pattiya Gatechumpol, Chawiporn Muktabhant, Patarachai Sereerat
Abstract: This study presents a deep learning system for breast cancer detection in mammography, developed using a modified EfficientNetV2 architecture with enhanced attention mechanisms. The model was trained on mammograms from a major Thai medical center and validated on three distinct datasets: an in-domain test set (9,421 cases), a biopsy-confirmed set (883 cases), and an out-of-domain generalizability set (761 cases) collected from two different hospitals. For cancer detection, the model achieved AUROCs of 0.89, 0.96, and 0.94 on the respective datasets. The system's lesion localization capability, evaluated using metrics including Lesion Localization Fraction (LLF) and Non-Lesion Localization Fraction (NLF), demonstrated robust performance in identifying suspicious regions. Clinical validation through concordance tests showed strong agreement with radiologists: 83.5% classification and 84.0% localization concordance for biopsy-confirmed cases, and 78.1% classification and 79.6% localization concordance for out-of-domain cases. Expert radiologists' acceptance rate also averaged 96.7% for biopsy-confirmed cases, and 89.3% for out-of-domain cases. The system achieved a System Usability Scale score of 74.17 for source hospital, and 69.20 for validation hospitals, indicating good clinical acceptance. These results demonstrate the model's effectiveness in assisting mammogram interpretation, with the potential to enhance breast cancer screening workflows in clinical practice.
Authors: Kurt Micallef, Claudia Borg
Abstract: Large Language Models (LLMs) have demonstrated remarkable performance across various Natural Language Processing (NLP) tasks, largely due to their generalisability and ability to perform tasks without additional training. However, their effectiveness for low-resource languages remains limited. In this study, we evaluate the performance of 55 publicly available LLMs on Maltese, a low-resource language, using a newly introduced benchmark covering 11 discriminative and generative tasks. Our experiments highlight that many models perform poorly, particularly on generative tasks, and that smaller fine-tuned models often perform better across all tasks. From our multidimensional analysis, we investigate various factors impacting performance. We conclude that prior exposure to Maltese during pre-training and instruction-tuning emerges as the most important factor. We also examine the trade-offs between fine-tuning and prompting, highlighting that while fine-tuning requires a higher initial cost, it yields better performance and lower inference costs. Through this work, we aim to highlight the need for more inclusive language technologies and recommend that researchers working with low-resource languages consider more "traditional" language modelling approaches.
Authors: Haokun Liu, Zhaoqi Ma, Yunong Li, Junichiro Sugihara, Yicheng Chen, Jinjie Li, Moju Zhao
Abstract: Heterogeneous multi-robot systems show great potential in complex tasks requiring hybrid cooperation. However, traditional approaches relying on static models often struggle with task diversity and dynamic environments. This highlights the need for generalizable intelligence that can bridge high-level reasoning with low-level execution across heterogeneous agents. To address this, we propose a hierarchical framework integrating a prompted Large Language Model (LLM) and a GridMask-enhanced fine-tuned Vision Language Model (VLM). The LLM decomposes tasks and constructs a global semantic map, while the VLM extracts task-specified semantic labels and 2D spatial information from aerial images to support local planning. Within this framework, the aerial robot follows an optimized global semantic path and continuously provides bird-view images, guiding the ground robot's local semantic navigation and manipulation, including target-absent scenarios where implicit alignment is maintained. Experiments on real-world cube or object arrangement tasks demonstrate the framework's adaptability and robustness in dynamic environments. To the best of our knowledge, this is the first demonstration of an aerial-ground heterogeneous system integrating VLM-based perception with LLM-driven task reasoning and motion planning.
Authors: Xinghang Li, Jingzhe Ding, Chao Peng, Bing Zhao, Xiang Gao, Hongwan Gao, Xinchen Gu
Abstract: The code generation capabilities of large language models(LLMs) have emerged as a critical dimension in evaluating their overall performance. However, prior research has largely overlooked the security risks inherent in the generated code. In this work, we introduce SafeGenBench, a benchmark specifically designed to assess the security of LLM-generated code. The dataset encompasses a wide range of common software development scenarios and vulnerability types. Building upon this benchmark, we develop an automatic evaluation framework that leverages both static application security testing(SAST) and LLM-based judging to assess the presence of security vulnerabilities in model-generated code. Through the empirical evaluation of state-of-the-art LLMs on SafeGenBench, we reveal notable deficiencies in their ability to produce vulnerability-free code. Our findings highlight pressing challenges and offer actionable insights for future advancements in the secure code generation performance of LLMs. The data and code will be released soon.
Authors: Tim Schneider, Guillaume Duret, Cristiana de Farias, Roberto Calandra, Liming Chen, Jan Peters
Abstract: Tactile perception has the potential to significantly enhance dexterous robotic manipulation by providing rich local information that can complement or substitute for other sensory modalities such as vision. However, because tactile sensing is inherently local, it is not well-suited for tasks that require broad spatial awareness or global scene understanding on its own. A human-inspired strategy to address this issue is to consider active perception techniques instead. That is, to actively guide sensors toward regions with more informative or significant features and integrate such information over time in order to understand a scene or complete a task. Both active perception and different methods for tactile sensing have received significant attention recently. Yet, despite advancements, both fields lack standardized benchmarks. To bridge this gap, we introduce the Tactile MNIST Benchmark Suite, an open-source, Gymnasium-compatible benchmark specifically designed for active tactile perception tasks, including localization, classification, and volume estimation. Our benchmark suite offers diverse simulation scenarios, from simple toy environments all the way to complex tactile perception tasks using vision-based tactile sensors. Furthermore, we also offer a comprehensive dataset comprising 13,500 synthetic 3D MNIST digit models and 153,600 real-world tactile samples collected from 600 3D printed digits. Using this dataset, we train a CycleGAN for realistic tactile simulation rendering. By providing standardized protocols and reproducible evaluation frameworks, our benchmark suite facilitates systematic progress in the fields of tactile sensing and active perception.
Authors: Dongyeop Lee, Kwanhee Lee, Jinseok Chung, Namhoon Lee
Abstract: Sparsifying neural networks often suffers from seemingly inevitable performance degradation, and it remains challenging to restore the original performance despite much recent progress. Motivated by recent studies in robust optimization, we aim to tackle this problem by finding subnetworks that are both sparse and flat at the same time. Specifically, we formulate pruning as a sparsity-constrained optimization problem where flatness is encouraged as an objective. We solve it explicitly via an augmented Lagrange dual approach and extend it further by proposing a generalized projection operation, resulting in novel pruning methods called SAFE and its extension, SAFE$^+$. Extensive evaluations on standard image classification and language modeling tasks reveal that SAFE consistently yields sparse networks with improved generalization performance, which compares competitively to well-established baselines. In addition, SAFE demonstrates resilience to noisy data, making it well-suited for real-world conditions.
Authors: Arun Sharma, Mingzhou Yang, Majid Farhadloo, Subhankar Ghosh, Bharat Jayaprakash, Shashi Shekhar
Abstract: Given trajectory data, a domain-specific study area, and a user-defined threshold, we aim to find anomalous trajectories indicative of possible GPS spoofing (e.g., fake trajectory). The problem is societally important to curb illegal activities in international waters, such as unauthorized fishing and illicit oil transfers. The problem is challenging due to advances in AI generated in deep fakes generation (e.g., additive noise, fake trajectories) and lack of adequate amount of labeled samples for ground-truth verification. Recent literature shows promising results for anomalous trajectory detection using generative models despite data sparsity. However, they do not consider fine-scale spatiotemporal dependencies and prior physical knowledge, resulting in higher false-positive rates. To address these limitations, we propose a physics-informed diffusion model that integrates kinematic constraints to identify trajectories that do not adhere to physical laws. Experimental results on real-world datasets in the maritime and urban domains show that the proposed framework results in higher prediction accuracy and lower estimation error rate for anomaly detection and trajectory generation methods, respectively. Our implementation is available at https://github.com/arunshar/Physics-Informed-Diffusion-Probabilistic-Model.
URLs: https://github.com/arunshar/Physics-Informed-Diffusion-Probabilistic-Model.
Authors: Sanjoy Chowdhury, Mohamed Elmoghany, Yohan Abeysinghe, Junjie Fei, Sayan Nag, Salman Khan, Mohamed Elhoseiny, Dinesh Manocha
Abstract: Large multimodal models (LMMs) have shown remarkable progress in audio-visual understanding, yet they struggle with real-world scenarios that require complex reasoning across extensive video collections. Existing benchmarks for video question answering remain limited in scope, typically involving one clip per query, which falls short of representing the challenges of large-scale, audio-visual retrieval and reasoning encountered in practical applications. To bridge this gap, we introduce a novel task named AV-HaystacksQA, where the goal is to identify salient segments across different videos in response to a query and link them together to generate the most informative answer. To this end, we present AVHaystacks, an audio-visual benchmark comprising 3100 annotated QA pairs designed to assess the capabilities of LMMs in multi-video retrieval and temporal grounding task. Additionally, we propose a model-agnostic, multi-agent framework MAGNET to address this challenge, achieving up to 89% and 65% relative improvements over baseline methods on BLEU@4 and GPT evaluation scores in QA task on our proposed AVHaystacks. To enable robust evaluation of multi-video retrieval and temporal grounding for optimal response generation, we introduce two new metrics, STEM, which captures alignment errors between a ground truth and a predicted step sequence and MTGS, to facilitate balanced and interpretable evaluation of segment-level grounding performance. Project: https://schowdhury671.github.io/magnet_project/
Authors: Yuya Okada, Takayuki Nishio
Abstract: We propose SALT (Split-Adaptive Lightweight Tuning), a lightweight model adaptation framework for Split Computing under closed constraints, where the head and tail networks are proprietary and inaccessible to users. In such closed environments, conventional adaptation methods are infeasible since they require access to model parameters or architectures. SALT addresses this challenge by introducing a compact, trainable adapter on the client side to refine latent features from the head network, enabling user-specific adaptation without modifying the original models or increasing communication overhead. We evaluate SALT on user-specific classification tasks with CIFAR-10 and CIFAR-100, demonstrating improved accuracy with lower training latency compared to fine-tuning methods. Furthermore, SALT facilitates model adaptation for robust inference over lossy networks, a common challenge in edge-cloud environments. With minimal deployment overhead, SALT offers a practical solution for personalized inference in edge AI systems under strict system constraints.
Authors: Jie He, Minglang Chen, Minying Lu, Bocheng Liang, Junming Wei, Guiyan Peng, Jiaxi Chen, Ying Tan
Abstract: Accurate ultrasound image segmentation is a prerequisite for precise biometrics and accurate assessment. Relying on manual delineation introduces significant errors and is time-consuming. However, existing segmentation models are designed based on objects in natural scenes, making them difficult to adapt to ultrasound objects with high noise and high similarity. This is particularly evident in small object segmentation, where a pronounced jagged effect occurs. Therefore, this paper proposes a fetal femur and cranial ultrasound image segmentation model based on feature perception and Mamba enhancement to address these challenges. Specifically, a longitudinal and transverse independent viewpoint scanning convolution block and a feature perception module were designed to enhance the ability to capture local detail information and improve the fusion of contextual information. Combined with the Mamba-optimized residual structure, this design suppresses the interference of raw noise and enhances local multi-dimensional scanning. The system builds global information and local feature dependencies, and is trained with a combination of different optimizers to achieve the optimal solution. After extensive experimental validation, the FAMSeg network achieved the fastest loss reduction and the best segmentation performance across images of varying sizes and orientations.
Authors: Shun Lei, Yaoxun Xu, Zhiwei Lin, Huaicheng Zhang, Wei Tan, Hangting Chen, Jianwei Yu, Yixuan Zhang, Chenyu Yang, Haina Zhu, Shuai Wang, Zhiyong Wu, Dong Yu
Abstract: Recent advances in large language models (LLMs) and audio language models have significantly improved music generation, particularly in lyrics-to-song generation. However, existing approaches still struggle with the complex composition of songs and the scarcity of high-quality data, leading to limitations in sound quality, musicality, instruction following, and vocal-instrument harmony. To address these challenges, we introduce LeVo, an LM-based framework consisting of LeLM and a music codec. LeLM is capable of parallelly modeling two types of tokens: mixed tokens, which represent the combined audio of vocals and accompaniment to achieve vocal-instrument harmony, and dual-track tokens, which separately encode vocals and accompaniment for high-quality song generation. It employs two decoder-only transformers and a modular extension training strategy to prevent interference between different token types. To further enhance musicality and instruction following, we introduce a multi-preference alignment method based on Direct Preference Optimization (DPO). This method handles diverse human preferences through a semi-automatic data construction process and DPO post-training. Experimental results demonstrate that LeVo consistently outperforms existing methods on both objective and subjective metrics. Ablation studies further justify the effectiveness of our designs. Audio examples are available at https://levo-demo.github.io/. Code is released at https://github.com/tencent-ailab/songgeneration.
URLs: https://levo-demo.github.io/., https://github.com/tencent-ailab/songgeneration.
Authors: Haoxiang Wang, Zinan Lin, Da Yu, Huishuai Zhang
Abstract: Generating high fidelity, differentially private (DP) synthetic images offers a promising route to share and analyze sensitive visual data without compromising individual privacy. However, existing DP image synthesis methods struggle to produce high resolution outputs that faithfully capture the structure of the original data. In this paper, we introduce a novel method, referred to as Synthesis via Private Textual Intermediaries (SPTI), that can generate high resolution DP images with easy adoption. The key idea is to shift the challenge of DP image synthesis from the image domain to the text domain by leveraging state of the art DP text generation methods. SPTI first summarizes each private image into a concise textual description using image to text models, then applies a modified Private Evolution algorithm to generate DP text, and finally reconstructs images using text to image models. Notably, SPTI requires no model training, only inference with off the shelf models. Given a private dataset, SPTI produces synthetic images of substantially higher quality than prior DP approaches. On the LSUN Bedroom dataset, SPTI attains an FID less than or equal to 26.71 under epsilon equal to 1.0, improving over Private Evolution FID of 40.36. Similarly, on MM CelebA HQ, SPTI achieves an FID less than or equal to 33.27 at epsilon equal to 1.0, compared to 57.01 from DP fine tuning baselines. Overall, our results demonstrate that Synthesis via Private Textual Intermediaries provides a resource efficient and proprietary model compatible framework for generating high resolution DP synthetic images, greatly expanding access to private visual datasets.
Authors: Jianhui Wei, Zikai Xiao, Danyu Sun, Luqi Gong, Zongxin Yang, Zuozhu Liu, Jian Wu
Abstract: Surgical video understanding is pivotal for enabling automated intraoperative decision-making, skill assessment, and postoperative quality improvement. However, progress in developing surgical video foundation models (FMs) remains hindered by the scarcity of large-scale, diverse datasets for pretraining and systematic evaluation. In this paper, we introduce \textbf{SurgBench}, a unified surgical video benchmarking framework comprising a pretraining dataset, \textbf{SurgBench-P}, and an evaluation benchmark, \textbf{SurgBench-E}. SurgBench offers extensive coverage of diverse surgical scenarios, with SurgBench-P encompassing 53 million frames across 22 surgical procedures and 11 specialties, and SurgBench-E providing robust evaluation across six categories (phase classification, camera motion, tool recognition, disease diagnosis, action classification, and organ detection) spanning 72 fine-grained tasks. Extensive experiments reveal that existing video FMs struggle to generalize across varied surgical video analysis tasks, whereas pretraining on SurgBench-P yields substantial performance improvements and superior cross-domain generalization to unseen procedures and modalities. Our dataset and code are available upon request.
Authors: Mirko Paolo Barbato, Giorgia Rigamonti, Davide Marelli, Paolo Napoletano
Abstract: Type 1 Diabetes (T1D) affects millions worldwide, requiring continuous monitoring to prevent severe hypo- and hyperglycemic events. While continuous glucose monitoring has improved blood glucose management, deploying predictive models on wearable devices remains challenging due to computational and memory constraints. To address this, we propose a novel Lightweight Sequential Transformer model designed for blood glucose prediction in T1D. By integrating the strengths of Transformers' attention mechanisms and the sequential processing of recurrent neural networks, our architecture captures long-term dependencies while maintaining computational efficiency. The model is optimized for deployment on resource-constrained edge devices and incorporates a balanced loss function to handle the inherent data imbalance in hypo- and hyperglycemic events. Experiments on two benchmark datasets, OhioT1DM and DiaTrend, demonstrate that the proposed model outperforms state-of-the-art methods in predicting glucose levels and detecting adverse events. This work fills the gap between high-performance modeling and practical deployment, providing a reliable and efficient T1D management solution.
Authors: Zeju Qiu, Simon Buchholz, Tim Z. Xiao, Maximilian Dax, Bernhard Sch\"olkopf, Weiyang Liu
Abstract: While large language models (LLMs) are driving the rapid advancement of artificial intelligence, effectively and reliably training these large models remains one of the field's most significant challenges. To address this challenge, we propose POET, a novel reParameterized training algorithm that uses Orthogonal Equivalence Transformation to optimize neurons. Specifically, POET reparameterizes each neuron with two learnable orthogonal matrices and a fixed random weight matrix. Because of its provable preservation of spectral properties of weight matrices, POET can stably optimize the objective function with improved generalization. We further develop efficient approximations that make POET flexible and scalable for training large-scale neural networks. Extensive experiments validate the effectiveness and scalability of POET in training LLMs.
Authors: Huixin Zhan, Jason H. Moore
Abstract: Surgeons exhibit distinct operating styles shaped by training, experience, and motor behavior-yet most surgical AI systems overlook this personalization signal. We propose a novel agentic modeling approach for surgeon-specific behavior prediction in robotic surgery, combining a discrete diffusion framework with a vision-language-action (VLA) pipeline. Gesture prediction is framed as a structured sequence denoising task, conditioned on multimodal inputs including surgical video, intent language, and personalized embeddings of surgeon identity and skill. These embeddings are encoded through natural language prompts using third-party language models, allowing the model to retain individual behavioral style without exposing explicit identity. We evaluate our method on the JIGSAWS dataset and demonstrate that it accurately reconstructs gesture sequences while learning meaningful motion fingerprints unique to each surgeon. To quantify the privacy implications of personalization, we perform membership inference attacks and find that more expressive embeddings improve task performance but simultaneously increase susceptibility to identity leakage. These findings demonstrate that while personalized embeddings improve performance, they also increase vulnerability to identity leakage, revealing the importance of balancing personalization with privacy risk in surgical modeling. Code is available at: https://github.com/huixin-zhan-ai/Surgeon_style_fingerprinting.
URLs: https://github.com/huixin-zhan-ai/Surgeon_style_fingerprinting.
Authors: Mansooreh Montazerin, Majd Al Aawar, Antonio Ortega, Ajitesh Srivastava
Abstract: Symbolic regression (SR) aims to discover closed-form mathematical expressions that accurately describe data, offering interpretability and analytical insight beyond standard black-box models. Existing SR methods often rely on population-based search or autoregressive modeling, which struggle with scalability and symbolic consistency. We introduce LIES (Logarithm, Identity, Exponential, Sine), a fixed neural network architecture with interpretable primitive activations that are optimized to model symbolic expressions. We develop a framework to extract compact formulae from LIES networks by training with an appropriate oversampling strategy and a tailored loss function to promote sparsity and to prevent gradient instability. After training, it applies additional pruning strategies to further simplify the learned expressions into compact formulae. Our experiments on SR benchmarks show that the LIES framework consistently produces sparse and accurate symbolic formulae outperforming all baselines. We also demonstrate the importance of each design component through ablation studies.
Authors: Arie Cattan, Alon Jacovi, Ori Ram, Jonathan Herzig, Roee Aharoni, Sasha Goldshtein, Eran Ofek, Idan Szpektor, Avi Caciularu
Abstract: Retrieval Augmented Generation (RAG) is a commonly used approach for enhancing large language models (LLMs) with relevant and up-to-date information. However, the retrieved sources can often contain conflicting information and it remains unclear how models should address such discrepancies. In this work, we first propose a novel taxonomy of knowledge conflict types in RAG, along with the desired model behavior for each type. We then introduce CONFLICTS, a high-quality benchmark with expert annotations of conflict types in a realistic RAG setting. CONFLICTS is the first benchmark that enables tracking progress on how models address a wide range of knowledge conflicts. We conduct extensive experiments on this benchmark, showing that LLMs often struggle to appropriately resolve conflicts between sources. While prompting LLMs to explicitly reason about the potential conflict in the retrieved documents significantly improves the quality and appropriateness of their responses, substantial room for improvement in future research remains.
Authors: Ananthu Aniraj, Cassio F. Dantas, Dino Ienco, Diego Marcos
Abstract: We introduce an attention-based method that uses learned binary attention masks to ensure that only attended image regions influence the prediction. Context can strongly affect object perception, sometimes leading to biased representations, particularly when objects appear in out-of-distribution backgrounds. At the same time, many image-level object-centric tasks require identifying relevant regions, often requiring context. To address this conundrum, we propose a two-stage framework: stage 1 processes the full image to discover object parts and identify task-relevant regions, while stage 2 leverages input attention masking to restrict its receptive field to these regions, enabling a focused analysis while filtering out potentially spurious information. Both stages are trained jointly, allowing stage 2 to refine stage 1. Extensive experiments across diverse benchmarks demonstrate that our approach significantly improves robustness against spurious correlations and out-of-distribution backgrounds. Code: https://github.com/ananthu-aniraj/ifam
Authors: Chunming He, Kai Li, Yachao Zhang, Ziyun Yang, Youwei Pang, Longxiang Tang, Chengyu Fang, Yulun Zhang, Linghe Kong, Xiu Li, Sina Farsiu
Abstract: Incompletely-Supervised Concealed Object Segmentation (ISCOS) involves segmenting objects that seamlessly blend into their surrounding environments, utilizing incompletely annotated data, such as weak and semi-annotations, for model training. This task remains highly challenging due to (1) the limited supervision provided by the incompletely annotated training data, and (2) the difficulty of distinguishing concealed objects from the background, which arises from the intrinsic similarities in concealed scenarios. In this paper, we introduce the first unified method for ISCOS to address these challenges. To tackle the issue of incomplete supervision, we propose a unified mean-teacher framework, SEE, that leverages the vision foundation model, ``\emph{Segment Anything Model (SAM)}'', to generate pseudo-labels using coarse masks produced by the teacher model as prompts. To mitigate the effect of low-quality segmentation masks, we introduce a series of strategies for pseudo-label generation, storage, and supervision. These strategies aim to produce informative pseudo-labels, store the best pseudo-labels generated, and select the most reliable components to guide the student model, thereby ensuring robust network training. Additionally, to tackle the issue of intrinsic similarity, we design a hybrid-granularity feature grouping module that groups features at different granularities and aggregates these results. By clustering similar features, this module promotes segmentation coherence, facilitating more complete segmentation for both single-object and multiple-object images. We validate the effectiveness of our approach across multiple ISCOS tasks, and experimental results demonstrate that our method achieves state-of-the-art performance. Furthermore, SEE can serve as a plug-and-play solution, enhancing the performance of existing models.
Authors: Alyssa Pinnock, Shakya Jayakody, Kawsher A Roxy, Md Rubel Ahmed
Abstract: This paper introduces EdgeProfiler, a fast profiling framework designed for evaluating lightweight Large Language Models (LLMs) on edge systems. While LLMs offer remarkable capabilities in natural language understanding and generation, their high computational, memory, and power requirements often confine them to cloud environments. EdgeProfiler addresses these challenges by providing a systematic methodology for assessing LLM performance in resource-constrained edge settings. The framework profiles compact LLMs, including TinyLLaMA, Gemma3.1B, Llama3.2-1B, and DeepSeek-r1-1.5B, using aggressive quantization techniques and strict memory constraints. Analytical modeling is used to estimate latency, FLOPs, and energy consumption. The profiling reveals that 4-bit quantization reduces model memory usage by approximately 60-70%, while maintaining accuracy within 2-5% of full-precision baselines. Inference speeds are observed to improve by 2-3x compared to FP16 baselines across various edge devices. Power modeling estimates a 35-50% reduction in energy consumption for INT4 configurations, enabling practical deployment on hardware such as Raspberry Pi 4/5 and Jetson Orin Nano Super. Our findings emphasize the importance of efficient profiling tailored to lightweight LLMs in edge environments, balancing accuracy, energy efficiency, and computational feasibility.
Authors: Vivien van Veldhuizen, Vanessa Botha, Chunyao Lu, Melis Erdal Cesur, Kevin Groot Lipman, Edwin D. de Jong, Hugo Horlings, Cl\'arisa I. Sanchez, Cees G. M. Snoek, Lodewyk Wessels, Ritse Mann, Eric Marcus, Jonas Teuwen
Abstract: Foundation models (FMs) are changing the way medical images are analyzed by learning from large collections of unlabeled data. Instead of relying on manually annotated examples, FMs are pre-trained to learn general-purpose visual features that can later be adapted to specific clinical tasks with little additional supervision. In this review, we examine how FMs are being developed and applied in pathology, radiology, and ophthalmology, drawing on evidence from over 150 studies. We explain the core components of FM pipelines, including model architectures, self-supervised learning methods, and strategies for downstream adaptation. We also review how FMs are being used in each imaging domain and compare design choices across applications. Finally, we discuss key challenges and open questions to guide future research.
Authors: Chaoyang Zhou, Shunyu Liu, Zengmao Wang, Di Wang, Rong-Cheng Tu, Bo Du, Dacheng Tao
Abstract: Reward models are critical for improving large language models (LLMs), particularly in reinforcement learning from human feedback (RLHF) or inference-time verification. Current reward modeling typically relies on scores of overall responses to learn the outcome rewards for the responses. However, since the response-level scores are coarse-grained supervision signals, the reward model struggles to identify the specific components within a response trajectory that truly correlate with the scores, leading to poor generalization on unseen responses. In this paper, we propose to leverage generation probabilities to establish reward consistency between processes in the response trajectory, which allows the response-level supervisory signal to propagate across processes, thereby providing additional fine-grained signals for reward learning. Building on analysis under the Bayesian framework, we develop an intra-trajectory consistency regularization to enforce that adjacent processes with higher next-token generation probability maintain more consistent rewards. We apply the proposed regularization to the advanced outcome reward model, improving its performance on RewardBench. Besides, we show that the reward model trained with the proposed regularization induces better DPO-aligned policies and achieves better best-of-N (BON) inference-time verification results. Our code is provided in https://github.com/chaoyang101/ICRM.
Authors: Sushant Mehta, Raj Dandekar, Rajat Dandekar, Sreedath Panat
Abstract: We present the first comprehensive study of latent multi-head attention (MLA) for small language models, revealing interesting efficiency-quality trade-offs. Training 30M-parameter GPT models on 100,000 synthetic stories, we benchmark three architectural variants: standard multi-head attention (MHA), MLA, and MLA with rotary positional embeddings (MLA+RoPE). Our key finding is that MLA+RoPE with half-rank latent dimensions (r = d/2) achieves a 45% KV-cache memory reduction while incurring only a 0.3% increase in validation loss (essentially matching MHA quality)- a Pareto improvement for memory constrained deployment. We further show that RoPE is crucial for MLA in small models: without it, MLA underperforms vanilla attention by 3-5%, but with RoPE, it surpasses vanilla by 2%. Inference benchmarks on NVIDIA A100 GPUs reveal that MLA with r=d/2 achieves a 1.4 times speedup over full-rank MLA while maintaining the memory savings. GPT-4 evaluations corroborate perplexity results, with ours achieving the highest quality scores (7.4/10) across grammar, creativity, and consistency metrics. Code and models will be released upon acceptance.
Authors: Jiaqi Tang, Yu Xia, Yi-Feng Wu, Yuwei Hu, Yuhui Chen, Qing-Guo Chen, Xiaogang Xu, Xiangyu Wu, Hao Lu, Yanqing Ma, Shiyin Lu, Qifeng Chen
Abstract: The advent of autonomous agents is transforming interactions with Graphical User Interfaces (GUIs) by employing natural language as a powerful intermediary. Despite the predominance of Supervised Fine-Tuning (SFT) methods in current GUI agents for achieving spatial localization, these methods face substantial challenges due to their limited capacity to accurately perceive positional data. Existing strategies, such as reinforcement learning, often fail to assess positional accuracy effectively, thereby restricting their utility. In response, we introduce Location Preference Optimization (LPO), a novel approach that leverages locational data to optimize interaction preferences. LPO uses information entropy to predict interaction positions by focusing on zones rich in information. Besides, it further introduces a dynamic location reward function based on physical distance, reflecting the varying importance of interaction positions. Supported by Group Relative Preference Optimization (GRPO), LPO facilitates an extensive exploration of GUI environments and significantly enhances interaction precision. Comprehensive experiments demonstrate LPO's superior performance, achieving SOTA results across both offline benchmarks and real-world online evaluations. Our code will be made publicly available soon, at https://github.com/AIDC-AI/LPO.
Authors: Christoph Schuhmann, Robert Kaczmarczyk, Gollam Rabby, Felix Friedrich, Maurice Kraus, Kourosh Nadi, Huu Nguyen, Kristian Kersting, S\"oren Auer
Abstract: The advancement of text-to-speech and audio generation models necessitates robust benchmarks for evaluating the emotional understanding capabilities of AI systems. Current speech emotion recognition (SER) datasets often exhibit limitations in emotional granularity, privacy concerns, or reliance on acted portrayals. This paper introduces EmoNet-Voice, a new resource for speech emotion detection, which includes EmoNet-Voice Big, a large-scale pre-training dataset (featuring over 4,500 hours of speech across 11 voices, 40 emotions, and 4 languages), and EmoNet-Voice Bench, a novel benchmark dataset with human expert annotations. EmoNet-Voice is designed to evaluate SER models on a fine-grained spectrum of 40 emotion categories with different levels of intensities. Leveraging state-of-the-art voice generation, we curated synthetic audio snippets simulating actors portraying scenes designed to evoke specific emotions. Crucially, we conducted rigorous validation by psychology experts who assigned perceived intensity labels. This synthetic, privacy-preserving approach allows for the inclusion of sensitive emotional states often absent in existing datasets. Lastly, we introduce Empathic Insight Voice models that set a new standard in speech emotion recognition with high agreement with human experts. Our evaluations across the current model landscape exhibit valuable findings, such as high-arousal emotions like anger being much easier to detect than low-arousal states like concentration.
Authors: Shangshang Wang, Julian Asilis, \"Omer Faruk Akg\"ul, Enes Burak Bilgin, Ollie Liu, Deqing Fu, Willie Neiswanger
Abstract: How cost-effectively can we elicit strong reasoning in language models by leveraging their underlying representations? We answer this question with Resa, a family of 1.5B reasoning models trained via a novel and efficient sparse autoencoder tuning (SAE-Tuning) procedure. This method first trains an SAE to capture reasoning abilities from a source model, and then uses the trained SAE to guide a standard supervised fine-tuning process to elicit such abilities in a target model, all using verified question-answer data without any reasoning traces. Notably, when applied to certain base models before further RL post-training, SAE-Tuning retains >97% of its RL-trained counterpart's reasoning performance while reducing training costs by >2000x to roughly \$1 and training time by >450x to around 20 minutes. Furthermore, when applied to lightly RL-trained models (e.g., within 1 hour on 2 GPUs), it enables reasoning performance such as 43.33% Pass@1 on AIME24 and 90% Pass@1 on AMC23 for only around \$1 additional cost. Surprisingly, the reasoning abilities extracted via SAEs are potentially both generalizable and modular. Generality means abilities extracted from one dataset still elevate performance on a larger and overlapping corpus. Modularity means abilities extracted from Qwen or Qwen-Math can be attached to the R1-Distill model at test time, without any retraining, and yield comparable gains. Extensive ablations validate these findings and all artifacts are fully open-sourced.
Authors: Longzhen Han, Awes Mubarak, Almas Baimagambetov, Nikolaos Polatidis, Thar Baker
Abstract: Multimodal Large Language Models (MLLMs) have rapidly evolved beyond text generation, now spanning diverse output modalities including images, music, video, human motion, and 3D objects, by integrating language with other sensory modalities under unified architectures. This survey categorises six primary generative modalities and examines how foundational techniques, namely Self-Supervised Learning (SSL), Mixture of Experts (MoE), Reinforcement Learning from Human Feedback (RLHF), and Chain-of-Thought (CoT) prompting, enable cross-modal capabilities. We analyze key models, architectural trends, and emergent cross-modal synergies, while highlighting transferable techniques and unresolved challenges. Architectural innovations like transformers and diffusion models underpin this convergence, enabling cross-modal transfer and modular specialization. We highlight emerging patterns of synergy, and identify open challenges in evaluation, modularity, and structured reasoning. This survey offers a unified perspective on MLLM development and identifies critical paths toward more general-purpose, adaptive, and interpretable multimodal systems.
Authors: Saswat Das, Jameson Sandler, Ferdinando Fioretto
Abstract: Large Language Model agents have begun to appear as personal assistants, customer service bots, and clinical aides. While these applications deliver substantial operational benefits, they also require continuous access to sensitive data, which increases the likelihood of unauthorized disclosures. This study proposes an auditing framework for conversational privacy that quantifies and audits these risks. The proposed Conversational Manipulation for Privacy Leakage (CMPL) framework, is an iterative probing strategy designed to stress-test agents that enforce strict privacy directives. Rather than focusing solely on a single disclosure event, CMPL simulates realistic multi-turn interactions to systematically uncover latent vulnerabilities. Our evaluation on diverse domains, data modalities, and safety configurations demonstrate the auditing framework's ability to reveal privacy risks that are not deterred by existing single-turn defenses. In addition to introducing CMPL as a diagnostic tool, the paper delivers (1) an auditing procedure grounded in quantifiable risk metrics and (2) an open benchmark for evaluation of conversational privacy across agent implementations.
Authors: Andrea Gaggioli, Sabrina Bartolotta, Andrea Ubaldi, Katusha Gerardini, Eleonora Diletta Sarcinella, Alice Chirico
Abstract: Artificial Intelligence holds significant potential to enhance human creativity. However, achieving this vision requires a clearer understanding of how such enhancement can be effectively realized. Drawing on a relational and distributed cognition perspective, we identify three fundamental modes by which AI can support and shape creative processes: Support, where AI acts as a tool; Synergy, where AI and humans collaborate in complementary ways; and Symbiosis, where human and AI cognition become so integrated that they form a unified creative system. These modes are defined along two key dimensions: the level of technical autonomy exhibited by the AI system (i.e., its ability to operate independently and make decisions without human intervention), and the degree of perceived agency attributed to it (i.e., the extent to which the AI is experienced as an intentional or creative partner). We examine how each configuration influences different levels of creativity from everyday problem solving to paradigm shifting innovation and discuss the implications for ethics, research, and the design of future human AI creative systems.
Authors: Felix H\"arer
Abstract: Recent advancements in LLMs indicate potential for novel applications, e.g., through reasoning capabilities in the latest OpenAI and DeepSeek models. For applying these models in specific domains beyond text generation, LLM-based multi-agent approaches can be utilized that solve complex tasks by combining reasoning techniques, code generation, and software execution. Applications might utilize these capabilities and the knowledge of specialized LLM agents. However, while many evaluations are performed on LLMs, reasoning techniques, and applications individually, their joint specification and combined application is not explored well. Defined specifications for multi-agent LLM systems are required to explore their potential and their suitability for specific applications, allowing for systematic evaluations of LLMs, reasoning techniques, and related aspects. This paper reports the results of exploratory research to specify and evaluate these aspects through a multi-agent system. The system architecture and prototype are extended from previous research and a specification is introduced for multi-agent systems. Test cases involving cybersecurity tasks indicate feasibility of the architecture and evaluation approach. In particular, the results show the evaluation of question answering, server security, and network security tasks that were completed correctly by agents with LLMs from OpenAI and DeepSeek.
Authors: Lang Yin, Debangshu Banerjee, Gagandeep Singh
Abstract: Chain of Thought (CoT) has been applied to various large language models (LLMs) and proven to be effective in improving the quality of outputs. In recent studies, transformers are proven to have absolute upper bounds in terms of expressive power, and consequently, they cannot solve many computationally difficult problems. However, empowered by CoT, transformers are proven to be able to solve some difficult problems effectively, such as the $k$-parity problem. Nevertheless, those works rely on two imperative assumptions: (1) identical training and testing distribution, and (2) corruption-free training data with correct reasoning steps. However, in the real world, these assumptions do not always hold. Although the risks of data shifts have caught attention, our work is the first to rigorously study the exact harm caused by such shifts to the best of our knowledge. Focusing on the $k$-parity problem, in this work we investigate the joint impact of two types of data shifts: the distribution shifts and data poisoning, on the quality of trained models obtained by a well-established CoT decomposition. In addition to revealing a surprising phenomenon that CoT leads to worse performance on learning parity than directly generating the prediction, our technical results also give a rigorous and comprehensive explanation of the mechanistic reasons of such impact.
Authors: Huaying Yuan, Zheng Liu, Junjie Zhou, Hongjin Qian, Ji-Rong Wen, Zhicheng Dou
Abstract: Long video understanding (LVU) presents a significant challenge for current multi-modal large language models (MLLMs) due to the task's inherent complexity and context window constraint. It is widely assumed that addressing LVU tasks requires foundation MLLMs with extended context windows, strong visual perception capabilities, and proficient domain expertise. In this work, we challenge this common belief by introducing VideoDeepResearch, a novel agentic framework for long video understanding. Our approach relies solely on a text-only large reasoning model (LRM) combined with a modular multi-modal toolkit, including multimodal retrievers and visual perceivers, all of which are readily available in practice. For each LVU task, the system formulates a problem-solving strategy through reasoning, while selectively accessing and utilizing essential video content via tool using. We conduct extensive experiments on popular LVU benchmarks, including MLVU, Video-MME, and LVBench. Our results demonstrate that VideoDeepResearch achieves substantial improvements over existing MLLM baselines, surpassing the previous state-of-the-art by 9.6%, 6.6%, and 3.9% on MLVU (test), LVBench, and LongVideoBench, respectively. These findings highlight the promise of agentic systems in overcoming key challenges in LVU problems.
Authors: Chen Yueh-Han, Nitish Joshi, Yulin Chen, Maksym Andriushchenko, Rico Angell, He He
Abstract: Current LLM safety defenses fail under decomposition attacks, where a malicious goal is decomposed into benign subtasks that circumvent refusals. The challenge lies in the existing shallow safety alignment techniques: they only detect harm in the immediate prompt and do not reason about long-range intent, leaving them blind to malicious intent that emerges over a sequence of seemingly benign instructions. We therefore propose adding an external monitor that observes the conversation at a higher granularity. To facilitate our study of monitoring decomposition attacks, we curate the largest and most diverse dataset to date, including question-answering, text-to-image, and agentic tasks. We verify our datasets by testing them on frontier LLMs and show an 87% attack success rate on average on GPT-4o. This confirms that decomposition attack is broadly effective. Additionally, we find that random tasks can be injected into the decomposed subtasks to further obfuscate malicious intents. To defend in real time, we propose a lightweight sequential monitoring framework that cumulatively evaluates each subtask. We show that a carefully prompt engineered lightweight monitor achieves a 93% defense success rate, beating reasoning models like o3 mini as a monitor. Moreover, it remains robust against random task injection and cuts cost by 90% and latency by 50%. Our findings suggest that lightweight sequential monitors are highly effective in mitigating decomposition attacks and are viable in deployment.
Authors: Houyi Li, Wenzhen Zheng, Qiufeng Wang, Zhenyu Ding, Haoying Wang, Zili Wang, Shijie Xuyang, Ning Ding, Shuigeng Zhou, Xiangyu Zhang, Daxin Jiang
Abstract: Training Large Language Models (LLMs) is prohibitively expensive, creating a critical scaling gap where insights from small-scale experiments often fail to transfer to resource-intensive production systems, thereby hindering efficient innovation. To bridge this, we introduce Farseer, a novel and refined scaling law offering enhanced predictive accuracy across scales. By systematically constructing a model loss surface $L(N,D)$, Farseer achieves a significantly better fit to empirical data than prior laws (e.g., Chinchilla's law). Our methodology yields accurate, robust, and highly generalizable predictions, demonstrating excellent extrapolation capabilities, improving upon Chinchilla's law by reducing extrapolation error by 433\%. This allows for the reliable evaluation of competing training strategies across all $(N,D)$ settings, enabling conclusions from small-scale ablation studies to be confidently extrapolated to predict large-scale performance. Furthermore, Farseer provides new insights into optimal compute allocation, better reflecting the nuanced demands of modern LLM training. To validate our approach, we trained an extensive suite of approximately 1,000 LLMs across diverse scales and configurations, consuming roughly 3 million NVIDIA H100 GPU hours. We are comprehensively open-sourcing all models, data, results, and logs at https://github.com/Farseer-Scaling-Law/Farseer to foster further research.
Authors: Federico Pennino, Bianca Raimondi, Massimo Rondelli, Andrea Gurioli, Maurizio Gabbrielli
Abstract: Generating accurate and executable code using large language models (LLMs) is challenging for languages with limited public training data compared to popular languages such as Python. This paper introduces a generalizable approach that uses small-scale code versions of the Qwen 2.5 model combined with Group Relative Policy Optimization (GRPO) to enable effective code generation through explicit reasoning steps, which is particularly beneficial for languages with smaller source code databases. Using Prolog as a representative use case -- given its limited online presence -- the initial model faced challenges in generating executable code. After some training steps, the model successfully produces logically consistent and syntactically accurate code by directly integrating reasoning-driven feedback into the reinforcement learning loop. Experimental evaluations using mathematical logic problem benchmarks illustrate significant improvements in reasoning quality, code accuracy, and logical correctness, underscoring the potential of this approach to benefit a wide range of programming languages lacking extensive training resources.
Authors: Zoher Kachwala, Danishjeet Singh, Danielle Yang, Filippo Menczer
Abstract: As image generators produce increasingly realistic images, concerns about potential misuse continue to grow. Supervised detection relies on large, curated datasets and struggles to generalize across diverse generators. In this work, we investigate the use of pre-trained Vision-Language Models (VLMs) for zero-shot detection of AI-generated images. While off-the-shelf VLMs exhibit some task-specific reasoning and chain-of-thought prompting offers gains, we show that task-aligned prompting elicits more focused reasoning and significantly improves performance without fine-tuning. Specifically, prefixing the model's response with the phrase "Let's examine the style and the synthesis artifacts" -- a method we call zero-shot-s$^2$ -- boosts Macro F1 scores by 8%-29%. These gains are consistent for two widely used open-source models and across three recent, diverse datasets spanning human faces, objects, and animals with images generated by 16 different models -- demonstrating strong generalization. We further evaluate the approach across three additional model sizes and observe improvements in most dataset-model combinations -- suggesting robustness to model scale. Surprisingly, self-consistency, a behavior previously observed in language reasoning, where aggregating answers from diverse reasoning paths improves performance, also holds in this setting. Even here, zero-shot-s$^2$ scales better than chain-of-thought in most cases -- indicating that it elicits more useful diversity. Our findings show that task-aligned prompts elicit more focused reasoning and enhance latent capabilities in VLMs, like the detection of AI-generated images -- offering a simple, generalizable, and explainable alternative to supervised methods. Our code is publicly available on github: https://github.com/Zoher15/Zero-shot-s2.
Authors: Uttej Kallakurik, Edward Humes, Rithvik Jonna, Xiaomin Lin, Tinoosh Mohsenin
Abstract: Large Language Models (LLMs) have significant impact on the healthcare scenarios but remain prohibitively large for deployment in real-time, resource-constrained environments such as edge devices. In this work, we introduce a novel medical assistant system, optimized through our general-purpose compression framework, which tailors Large Language Models (LLMs) for deployment in specialized domains. By measuring neuron saliency on domain-specific data, our method can aggressively prune irrelevant neurons, reducing model size while preserving performance. Following pruning, we apply post-training quantization to further reduce the memory footprint, and evaluate the compressed model across medical benchmarks including MedMCQA, MedQA, and PubMedQA. We also deploy the 50\% compressed Gemma and the 67\% compressed LLaMA3 models on Jetson Orin Nano (18.7W peak) and Raspberry Pi 5 (6.3W peak), achieving real-time, energy-efficient inference under hardware constraints.
Authors: Cheng-Kang Chou, Chan-Jan Hsu, Ho-Lam Chung, Liang-Hsuan Tseng, Hsi-Chun Cheng, Yu-Kuan Fu, Kuan Po Huang, Hung-Yi Lee
Abstract: We propose a self-refining framework that enhances ASR performance with only unlabeled datasets. The process starts with an existing ASR model generating pseudo-labels on unannotated speech, which are then used to train a high-fidelity text-to-speech (TTS) system. Then, synthesized speech text pairs are bootstrapped into the original ASR system, completing the closed-loop self-improvement cycle. We demonstrated the effectiveness of the framework on Taiwanese Mandarin speech. Leveraging 6,000 hours of unlabeled speech, a moderate amount of text data, and synthetic content from the AI models, we adapt Whisper-large-v2 into a specialized model, Twister. Twister reduces error rates by up to 20% on Mandarin and 50% on Mandarin-English code-switching benchmarks compared to Whisper. Results highlight the framework as a compelling alternative to pseudo-labeling self-distillation approaches and provides a practical pathway for improving ASR performance in low-resource or domain-specific settings.
Authors: Jin Kim, Muhammad Wahi-Anwa, Sangyun Park, Shawn Shin, John M. Hoffman, Matthew S. Brown
Abstract: Agentic Artificial Intelligence (AI) systems leveraging Large Language Models (LLMs) exhibit significant potential for complex reasoning, planning, and tool utilization. We demonstrate that a specialized computer vision system can be built autonomously from a natural language prompt using Agentic AI methods. This involved extending SimpleMind (SM), an open-source Cognitive AI environment with configurable tools for medical image analysis, with an LLM-based agent, implemented using OpenManus, to automate the planning (tool configuration) for a particular computer vision task. We provide a proof-of-concept demonstration that an agentic system can interpret a computer vision task prompt, plan a corresponding SimpleMind workflow by decomposing the task and configuring appropriate tools. From the user input prompt, "provide sm (SimpleMind) config for lungs, heart, and ribs segmentation for cxr (chest x-ray)"), the agent LLM was able to generate the plan (tool configuration file in YAML format), and execute SM-Learn (training) and SM-Think (inference) scripts autonomously. The computer vision agent automatically configured, trained, and tested itself on 50 chest x-ray images, achieving mean dice scores of 0.96, 0.82, 0.83, for lungs, heart, and ribs, respectively. This work shows the potential for autonomous planning and tool configuration that has traditionally been performed by a data scientist in the development of computer vision applications.
Authors: Xiaoyu Ma, Hao Chen, Yongjian Deng
Abstract: Different modalities hold considerable gaps in optimization trajectories, including speeds and paths, which lead to modality laziness and modality clash when jointly training multimodal models, resulting in insufficient and imbalanced multimodal learning. Existing methods focus on enforcing the weak modality by adding modality-specific optimization objectives, aligning their optimization speeds, or decomposing multimodal learning to enhance unimodal learning. These methods fail to achieve both unimodal sufficiency and multimodal balance. In this paper, we, for the first time, address both concerns by proposing multimodal Data Remixing, including decoupling multimodal data and filtering hard samples for each modality to mitigate modality imbalance; and then batch-level reassembling to align the gradient directions and avoid cross-modal interference, thus enhancing unimodal learning sufficiency. Experimental results demonstrate that our method can be seamlessly integrated with existing approaches, improving accuracy by approximately 6.50%$\uparrow$ on CREMAD and 3.41%$\uparrow$ on Kinetic-Sounds, without training set expansion or additional computational overhead during inference. The source code is available at https://github.com/MatthewMaxy/Remix_ICML2025.
Authors: Jiacong Wang, Zijian Kang, Haochen Wang, Haiyong Jiang, Jiawen Li, Bohong Wu, Ya Wang, Jiao Ran, Xiao Liang, Chao Feng, Jun Xiao
Abstract: In the field of multimodal chain-of-thought (CoT) reasoning, existing approaches predominantly rely on reasoning on pure language space, which inherently suffers from language bias and is largely confined to math or science domains. This narrow focus limits their ability to handle complex visual reasoning tasks that demand comprehensive understanding of image details. To address these limitations, this paper introduces VGR, a novel reasoning multimodal large language model (MLLM) with enhanced fine-grained visual perception capabilities. Unlike traditional MLLMs that answer the question or reasoning solely on the language space, our VGR first detects relevant regions that may help to solve problems, and then provides precise answers based on replayed image regions. To achieve this, we conduct a large-scale SFT dataset called VGR -SFT that contains reasoning data with mixed vision grounding and language deduction. The inference pipeline of VGR allows the model to choose bounding boxes for visual reference and a replay stage is introduced to integrates the corresponding regions into the reasoning process, enhancing multimodel comprehension. Experiments on the LLaVA-NeXT-7B baseline show that VGR achieves superior performance on multi-modal benchmarks requiring comprehensive image detail understanding. Compared to the baseline, VGR uses only 30\% of the image token count while delivering scores of +4.1 on MMStar, +7.1 on AI2D, and a +12.9 improvement on ChartQA.